xref: /netbsd-src/lib/libc/arch/aarch64/softfloat/softfloat.h (revision 138c67b5ef586c825b0cabe69fa1e7b0d5d077ab)
1 /* $NetBSD: softfloat.h,v 1.2 2018/12/06 19:17:13 tih Exp $ */
2 
3 /* This is a derivative work. */
4 
5 /*
6 ===============================================================================
7 
8 This C header file is part of the SoftFloat IEC/IEEE Floating-point
9 Arithmetic Package, Release 2a.
10 
11 Written by John R. Hauser.  This work was made possible in part by the
12 International Computer Science Institute, located at Suite 600, 1947 Center
13 Street, Berkeley, California 94704.  Funding was partially provided by the
14 National Science Foundation under grant MIP-9311980.  The original version
15 of this code was written as part of a project to build a fixed-point vector
16 processor in collaboration with the University of California at Berkeley,
17 overseen by Profs. Nelson Morgan and John Wawrzynek.  More information
18 is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
19 arithmetic/SoftFloat.html'.
20 
21 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE.  Although reasonable effort
22 has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
23 TIMES RESULT IN INCORRECT BEHAVIOR.  USE OF THIS SOFTWARE IS RESTRICTED TO
24 PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
25 AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
26 
27 Derivative works are acceptable, even for commercial purposes, so long as
28 (1) they include prominent notice that the work is derivative, and (2) they
29 include prominent notice akin to these four paragraphs for those parts of
30 this code that are retained.
31 
32 ===============================================================================
33 */
34 
35 /*
36 -------------------------------------------------------------------------------
37 The macro `FLOATX80' must be defined to enable the extended double-precision
38 floating-point format `floatx80'.  If this macro is not defined, the
39 `floatx80' type will not be defined, and none of the functions that either
40 input or output the `floatx80' type will be defined.  The same applies to
41 the `FLOAT128' macro and the quadruple-precision format `float128'.
42 -------------------------------------------------------------------------------
43 */
44 /* #define FLOATX80 */
45 #define FLOAT128
46 
47 #include "softfloat-qp.h"
48 
49 #include <sys/endian.h>
50 #include <machine/ieeefp.h>
51 
52 /*
53 -------------------------------------------------------------------------------
54 Software IEC/IEEE floating-point types.
55 -------------------------------------------------------------------------------
56 */
57 typedef unsigned int float32;
58 typedef unsigned long long float64;
59 #ifdef FLOATX80
60 typedef struct {
61     unsigned short high;
62     unsigned long long low;
63 } floatx80;
64 #endif
65 #ifdef FLOAT128
66 typedef struct {
67 #if _BYTE_ORDER == _BIG_ENDIAN
68     unsigned long long high, low;
69 #else
70     unsigned long long low, high;
71 #endif
72 } float128;
73 #endif
74 
75 /*
76 -------------------------------------------------------------------------------
77 Software IEC/IEEE floating-point underflow tininess-detection mode.
78 -------------------------------------------------------------------------------
79 */
80 #ifndef SOFTFLOAT_FOR_GCC
81 extern int8 float_detect_tininess;
82 #endif
83 enum {
84     float_tininess_after_rounding  = 0,
85     float_tininess_before_rounding = 1
86 };
87 
88 /*
89 -------------------------------------------------------------------------------
90 Software IEC/IEEE floating-point rounding mode.
91 -------------------------------------------------------------------------------
92 */
93 extern fp_rnd float_rounding_mode;
94 #define float_round_nearest_even FP_RN
95 #define float_round_to_zero      FP_RZ
96 #define float_round_down         FP_RM
97 #define float_round_up           FP_RP
98 
99 /*
100 -------------------------------------------------------------------------------
101 Software IEC/IEEE floating-point exception flags.
102 -------------------------------------------------------------------------------
103 */
104 extern fp_except float_exception_flags;
105 extern fp_except float_exception_mask;
106 enum {
107     float_flag_inexact   = FP_X_IMP,
108     float_flag_underflow = FP_X_UFL,
109     float_flag_overflow  = FP_X_OFL,
110     float_flag_divbyzero = FP_X_DZ,
111     float_flag_invalid   = FP_X_INV
112 };
113 
114 /*
115 -------------------------------------------------------------------------------
116 Routine to raise any or all of the software IEC/IEEE floating-point
117 exception flags.
118 -------------------------------------------------------------------------------
119 */
120 void float_raise( fp_except );
121 
122 /*
123 -------------------------------------------------------------------------------
124 Software IEC/IEEE integer-to-floating-point conversion routines.
125 -------------------------------------------------------------------------------
126 */
127 float32 int32_to_float32( int32 );
128 float32 uint32_to_float32( uint32 );
129 float64 int32_to_float64( int32 );
130 float64 uint32_to_float64( uint32 );
131 #ifdef FLOATX80
132 floatx80 int32_to_floatx80( int32 );
133 floatx80 uint32_to_floatx80( uint32 );
134 #endif
135 #ifdef FLOAT128
136 float128 int32_to_float128( int32 );
137 float128 uint32_to_float128( uint32 );
138 #endif
139 float32 int64_to_float32( long long );
140 float64 int64_to_float64( long long );
141 #ifdef FLOATX80
142 floatx80 int64_to_floatx80( long long );
143 #endif
144 #ifdef FLOAT128
145 float128 int64_to_float128( long long );
146 #endif
147 
148 /*
149 -------------------------------------------------------------------------------
150 Software IEC/IEEE single-precision conversion routines.
151 -------------------------------------------------------------------------------
152 */
153 int float32_to_int32( float32 );
154 int float32_to_int32_round_to_zero( float32 );
155 unsigned int float32_to_uint32_round_to_zero( float32 );
156 long long float32_to_int64( float32 );
157 long long float32_to_int64_round_to_zero( float32 );
158 float64 float32_to_float64( float32 );
159 #ifdef FLOATX80
160 floatx80 float32_to_floatx80( float32 );
161 #endif
162 #ifdef FLOAT128
163 float128 float32_to_float128( float32 );
164 #endif
165 
166 /*
167 -------------------------------------------------------------------------------
168 Software IEC/IEEE single-precision operations.
169 -------------------------------------------------------------------------------
170 */
171 float32 float32_round_to_int( float32 );
172 float32 float32_add( float32, float32 );
173 float32 float32_sub( float32, float32 );
174 float32 float32_mul( float32, float32 );
175 float32 float32_div( float32, float32 );
176 float32 float32_rem( float32, float32 );
177 float32 float32_sqrt( float32 );
178 flag float32_eq( float32, float32 );
179 flag float32_le( float32, float32 );
180 flag float32_lt( float32, float32 );
181 flag float32_eq_signaling( float32, float32 );
182 flag float32_le_quiet( float32, float32 );
183 flag float32_lt_quiet( float32, float32 );
184 flag float32_is_signaling_nan( float32 );
185 
186 /*
187 -------------------------------------------------------------------------------
188 Software IEC/IEEE double-precision conversion routines.
189 -------------------------------------------------------------------------------
190 */
191 int float64_to_int32( float64 );
192 int float64_to_int32_round_to_zero( float64 );
193 unsigned int float64_to_uint32_round_to_zero( float64 );
194 long long float64_to_int64( float64 );
195 long long float64_to_int64_round_to_zero( float64 );
196 float32 float64_to_float32( float64 );
197 #ifdef FLOATX80
198 floatx80 float64_to_floatx80( float64 );
199 #endif
200 #ifdef FLOAT128
201 float128 float64_to_float128( float64 );
202 #endif
203 
204 /*
205 -------------------------------------------------------------------------------
206 Software IEC/IEEE double-precision operations.
207 -------------------------------------------------------------------------------
208 */
209 float64 float64_round_to_int( float64 );
210 float64 float64_add( float64, float64 );
211 float64 float64_sub( float64, float64 );
212 float64 float64_mul( float64, float64 );
213 float64 float64_div( float64, float64 );
214 float64 float64_rem( float64, float64 );
215 float64 float64_sqrt( float64 );
216 flag float64_eq( float64, float64 );
217 flag float64_le( float64, float64 );
218 flag float64_lt( float64, float64 );
219 flag float64_eq_signaling( float64, float64 );
220 flag float64_le_quiet( float64, float64 );
221 flag float64_lt_quiet( float64, float64 );
222 flag float64_is_signaling_nan( float64 );
223 
224 #ifdef FLOATX80
225 
226 /*
227 -------------------------------------------------------------------------------
228 Software IEC/IEEE extended double-precision conversion routines.
229 -------------------------------------------------------------------------------
230 */
231 int floatx80_to_int32( floatx80 );
232 int floatx80_to_int32_round_to_zero( floatx80 );
233 long long floatx80_to_int64( floatx80 );
234 long long floatx80_to_int64_round_to_zero( floatx80 );
235 float32 floatx80_to_float32( floatx80 );
236 float64 floatx80_to_float64( floatx80 );
237 #ifdef FLOAT128
238 float128 floatx80_to_float128( floatx80 );
239 #endif
240 
241 /*
242 -------------------------------------------------------------------------------
243 Software IEC/IEEE extended double-precision rounding precision.  Valid
244 values are 32, 64, and 80.
245 -------------------------------------------------------------------------------
246 */
247 extern int floatx80_rounding_precision;
248 
249 /*
250 -------------------------------------------------------------------------------
251 Software IEC/IEEE extended double-precision operations.
252 -------------------------------------------------------------------------------
253 */
254 floatx80 floatx80_round_to_int( floatx80 );
255 floatx80 floatx80_add( floatx80, floatx80 );
256 floatx80 floatx80_sub( floatx80, floatx80 );
257 floatx80 floatx80_mul( floatx80, floatx80 );
258 floatx80 floatx80_div( floatx80, floatx80 );
259 floatx80 floatx80_rem( floatx80, floatx80 );
260 floatx80 floatx80_sqrt( floatx80 );
261 flag floatx80_eq( floatx80, floatx80 );
262 flag floatx80_le( floatx80, floatx80 );
263 flag floatx80_lt( floatx80, floatx80 );
264 flag floatx80_eq_signaling( floatx80, floatx80 );
265 flag floatx80_le_quiet( floatx80, floatx80 );
266 flag floatx80_lt_quiet( floatx80, floatx80 );
267 flag floatx80_is_signaling_nan( floatx80 );
268 
269 #endif
270 
271 #ifdef FLOAT128
272 
273 /*
274 -------------------------------------------------------------------------------
275 Software IEC/IEEE quadruple-precision conversion routines.
276 -------------------------------------------------------------------------------
277 */
278 int float128_to_int32( float128 );
279 int float128_to_int32_round_to_zero( float128 );
280 long long float128_to_int64( float128 );
281 long long float128_to_int64_round_to_zero( float128 );
282 unsigned long long float128_to_uint64_round_to_zero( float128 );
283 float32 float128_to_float32( float128 );
284 float64 float128_to_float64( float128 );
285 #ifdef FLOATX80
286 floatx80 float128_to_floatx80( float128 );
287 #endif
288 
289 /*
290 -------------------------------------------------------------------------------
291 Software IEC/IEEE quadruple-precision operations.
292 -------------------------------------------------------------------------------
293 */
294 float128 float128_round_to_int( float128 );
295 float128 float128_add( float128, float128 );
296 float128 float128_sub( float128, float128 );
297 float128 float128_mul( float128, float128 );
298 float128 float128_div( float128, float128 );
299 float128 float128_rem( float128, float128 );
300 float128 float128_sqrt( float128 );
301 flag float128_eq( float128, float128 );
302 flag float128_le( float128, float128 );
303 flag float128_lt( float128, float128 );
304 flag float128_eq_signaling( float128, float128 );
305 flag float128_le_quiet( float128, float128 );
306 flag float128_lt_quiet( float128, float128 );
307 flag float128_is_signaling_nan( float128 );
308 flag float128_is_nan( float128 );
309 
310 #endif
311