softfloat.h revision 1.5 1
2 /*
3 ===============================================================================
4
5 This C header file is part of the SoftFloat IEC/IEEE Floating-point
6 Arithmetic Package, Release 2a.
7
8 Written by John R. Hauser. This work was made possible in part by the
9 International Computer Science Institute, located at Suite 600, 1947 Center
10 Street, Berkeley, California 94704. Funding was partially provided by the
11 National Science Foundation under grant MIP-9311980. The original version
12 of this code was written as part of a project to build a fixed-point vector
13 processor in collaboration with the University of California at Berkeley,
14 overseen by Profs. Nelson Morgan and John Wawrzynek. More information
15 is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
16 arithmetic/SoftFloat.html'.
17
18 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort
19 has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
20 TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO
21 PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
22 AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
23
24 Derivative works are acceptable, even for commercial purposes, so long as
25 (1) they include prominent notice that the work is derivative, and (2) they
26 include prominent notice akin to these four paragraphs for those parts of
27 this code that are retained.
28
29 ===============================================================================
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 */
41 /* #define FLOATX80 */
42 /* #define FLOAT128 */
43
44 #include <machine/ieeefp.h>
45
46 /*
47 -------------------------------------------------------------------------------
48 Software IEC/IEEE floating-point types.
49 -------------------------------------------------------------------------------
50 */
51 typedef unsigned int float32;
52 typedef unsigned long long float64;
53 #ifdef FLOATX80
54 typedef struct {
55 unsigned short high;
56 unsigned long long low;
57 } floatx80;
58 #endif
59 #ifdef FLOAT128
60 typedef struct {
61 unsigned long long high, low;
62 } float128;
63 #endif
64
65 /*
66 -------------------------------------------------------------------------------
67 Software IEC/IEEE floating-point underflow tininess-detection mode.
68 -------------------------------------------------------------------------------
69 */
70 extern int float_detect_tininess;
71 enum {
72 float_tininess_after_rounding = 0,
73 float_tininess_before_rounding = 1
74 };
75
76 /*
77 -------------------------------------------------------------------------------
78 Software IEC/IEEE floating-point rounding mode.
79 -------------------------------------------------------------------------------
80 */
81 extern fp_rnd float_rounding_mode;
82 enum {
83 float_round_nearest_even = FP_RN,
84 float_round_to_zero = FP_RZ,
85 float_round_down = FP_RM,
86 float_round_up = FP_RP
87 };
88
89 /*
90 -------------------------------------------------------------------------------
91 Software IEC/IEEE floating-point exception flags.
92 -------------------------------------------------------------------------------
93 */
94 extern fp_except float_exception_flags;
95 enum {
96 float_flag_inexact = FP_X_IMP,
97 float_flag_underflow = FP_X_UFL,
98 float_flag_overflow = FP_X_OFL,
99 float_flag_divbyzero = FP_X_DZ,
100 float_flag_invalid = FP_X_INV
101 };
102
103 /*
104 -------------------------------------------------------------------------------
105 Routine to raise any or all of the software IEC/IEEE floating-point
106 exception flags.
107 -------------------------------------------------------------------------------
108 */
109 void float_raise( int );
110
111 /*
112 -------------------------------------------------------------------------------
113 Software IEC/IEEE integer-to-floating-point conversion routines.
114 -------------------------------------------------------------------------------
115 */
116 float32 int32_to_float32( int );
117 float64 int32_to_float64( int );
118 #ifdef FLOATX80
119 floatx80 int32_to_floatx80( int );
120 #endif
121 #ifdef FLOAT128
122 float128 int32_to_float128( int );
123 #endif
124 #ifndef SOFTFLOAT_FOR_GCC /* __floatdi?f is in libgcc2.c */
125 float32 int64_to_float32( long long );
126 float64 int64_to_float64( long long );
127 #ifdef FLOATX80
128 floatx80 int64_to_floatx80( long long );
129 #endif
130 #ifdef FLOAT128
131 float128 int64_to_float128( long long );
132 #endif
133 #endif
134
135 /*
136 -------------------------------------------------------------------------------
137 Software IEC/IEEE single-precision conversion routines.
138 -------------------------------------------------------------------------------
139 */
140 int float32_to_int32( float32 );
141 int float32_to_int32_round_to_zero( float32 );
142 #if defined(SOFTFLOAT_FOR_GCC) && defined(SOFTFLOAT_NEED_FIXUNS)
143 unsigned int float32_to_uint32_round_to_zero( float32 );
144 #endif
145 #ifndef SOFTFLOAT_FOR_GCC /* __fix?fdi provided by libgcc2.c */
146 long long float32_to_int64( float32 );
147 long long float32_to_int64_round_to_zero( float32 );
148 #endif
149 float64 float32_to_float64( float32 );
150 #ifdef FLOATX80
151 floatx80 float32_to_floatx80( float32 );
152 #endif
153 #ifdef FLOAT128
154 float128 float32_to_float128( float32 );
155 #endif
156
157 /*
158 -------------------------------------------------------------------------------
159 Software IEC/IEEE single-precision operations.
160 -------------------------------------------------------------------------------
161 */
162 float32 float32_round_to_int( float32 );
163 float32 float32_add( float32, float32 );
164 float32 float32_sub( float32, float32 );
165 float32 float32_mul( float32, float32 );
166 float32 float32_div( float32, float32 );
167 float32 float32_rem( float32, float32 );
168 float32 float32_sqrt( float32 );
169 int float32_eq( float32, float32 );
170 int float32_le( float32, float32 );
171 int float32_lt( float32, float32 );
172 int float32_eq_signaling( float32, float32 );
173 int float32_le_quiet( float32, float32 );
174 int float32_lt_quiet( float32, float32 );
175 #ifndef SOFTFLOAT_FOR_GCC
176 int float32_is_signaling_nan( float32 );
177 #endif
178
179 /*
180 -------------------------------------------------------------------------------
181 Software IEC/IEEE double-precision conversion routines.
182 -------------------------------------------------------------------------------
183 */
184 int float64_to_int32( float64 );
185 int float64_to_int32_round_to_zero( float64 );
186 #if defined(SOFTFLOAT_FOR_GCC) && defined(SOFTFLOAT_NEED_FIXUNS)
187 unsigned int float64_to_uint32_round_to_zero( float64 );
188 #endif
189 #ifndef SOFTFLOAT_FOR_GCC /* __fix?fdi provided by libgcc2.c */
190 long long float64_to_int64( float64 );
191 long long float64_to_int64_round_to_zero( float64 );
192 #endif
193 float32 float64_to_float32( float64 );
194 #ifdef FLOATX80
195 floatx80 float64_to_floatx80( float64 );
196 #endif
197 #ifdef FLOAT128
198 float128 float64_to_float128( float64 );
199 #endif
200
201 /*
202 -------------------------------------------------------------------------------
203 Software IEC/IEEE double-precision operations.
204 -------------------------------------------------------------------------------
205 */
206 float64 float64_round_to_int( float64 );
207 float64 float64_add( float64, float64 );
208 float64 float64_sub( float64, float64 );
209 float64 float64_mul( float64, float64 );
210 float64 float64_div( float64, float64 );
211 float64 float64_rem( float64, float64 );
212 float64 float64_sqrt( float64 );
213 int float64_eq( float64, float64 );
214 int float64_le( float64, float64 );
215 int float64_lt( float64, float64 );
216 int float64_eq_signaling( float64, float64 );
217 int float64_le_quiet( float64, float64 );
218 int float64_lt_quiet( float64, float64 );
219 #ifndef SOFTFLOAT_FOR_GCC
220 int float64_is_signaling_nan( float64 );
221 #endif
222
223 #ifdef FLOATX80
224
225 /*
226 -------------------------------------------------------------------------------
227 Software IEC/IEEE extended double-precision conversion routines.
228 -------------------------------------------------------------------------------
229 */
230 int floatx80_to_int32( floatx80 );
231 int floatx80_to_int32_round_to_zero( floatx80 );
232 long long floatx80_to_int64( floatx80 );
233 long long floatx80_to_int64_round_to_zero( floatx80 );
234 float32 floatx80_to_float32( floatx80 );
235 float64 floatx80_to_float64( floatx80 );
236 #ifdef FLOAT128
237 float128 floatx80_to_float128( floatx80 );
238 #endif
239
240 /*
241 -------------------------------------------------------------------------------
242 Software IEC/IEEE extended double-precision rounding precision. Valid
243 values are 32, 64, and 80.
244 -------------------------------------------------------------------------------
245 */
246 extern int floatx80_rounding_precision;
247
248 /*
249 -------------------------------------------------------------------------------
250 Software IEC/IEEE extended double-precision operations.
251 -------------------------------------------------------------------------------
252 */
253 floatx80 floatx80_round_to_int( floatx80 );
254 floatx80 floatx80_add( floatx80, floatx80 );
255 floatx80 floatx80_sub( floatx80, floatx80 );
256 floatx80 floatx80_mul( floatx80, floatx80 );
257 floatx80 floatx80_div( floatx80, floatx80 );
258 floatx80 floatx80_rem( floatx80, floatx80 );
259 floatx80 floatx80_sqrt( floatx80 );
260 int floatx80_eq( floatx80, floatx80 );
261 int floatx80_le( floatx80, floatx80 );
262 int floatx80_lt( floatx80, floatx80 );
263 int floatx80_eq_signaling( floatx80, floatx80 );
264 int floatx80_le_quiet( floatx80, floatx80 );
265 int floatx80_lt_quiet( floatx80, floatx80 );
266 int floatx80_is_signaling_nan( floatx80 );
267
268 #endif
269
270 #ifdef FLOAT128
271
272 /*
273 -------------------------------------------------------------------------------
274 Software IEC/IEEE quadruple-precision conversion routines.
275 -------------------------------------------------------------------------------
276 */
277 int float128_to_int32( float128 );
278 int float128_to_int32_round_to_zero( float128 );
279 long long float128_to_int64( float128 );
280 long long float128_to_int64_round_to_zero( float128 );
281 float32 float128_to_float32( float128 );
282 float64 float128_to_float64( float128 );
283 #ifdef FLOATX80
284 floatx80 float128_to_floatx80( float128 );
285 #endif
286
287 /*
288 -------------------------------------------------------------------------------
289 Software IEC/IEEE quadruple-precision operations.
290 -------------------------------------------------------------------------------
291 */
292 float128 float128_round_to_int( float128 );
293 float128 float128_add( float128, float128 );
294 float128 float128_sub( float128, float128 );
295 float128 float128_mul( float128, float128 );
296 float128 float128_div( float128, float128 );
297 float128 float128_rem( float128, float128 );
298 float128 float128_sqrt( float128 );
299 int float128_eq( float128, float128 );
300 int float128_le( float128, float128 );
301 int float128_lt( float128, float128 );
302 int float128_eq_signaling( float128, float128 );
303 int float128_le_quiet( float128, float128 );
304 int float128_lt_quiet( float128, float128 );
305 int float128_is_signaling_nan( float128 );
306
307 #endif
308
309