1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 /*
28 * Copyright 2011, Richard Lowe.
29 */
30
31 #ifndef _LIBM_INLINES_H
32 #define _LIBM_INLINES_H
33
34 #ifdef __GNUC__
35
36 #include <sys/types.h>
37 #include <sys/ieeefp.h>
38
39 #ifdef __cplusplus
40 extern "C" {
41 #endif
42
43 extern __GNU_INLINE enum fp_class_type
fp_classf(float f)44 fp_classf(float f)
45 {
46 enum fp_class_type ret;
47 int fint; /* scratch for f as int */
48 uint64_t tmp;
49
50 __asm__ __volatile__(
51 "fabss %3,%3\n\t"
52 "st %3,%1\n\t"
53 "ld %1,%0\n\t"
54 "orcc %%g0,%0,%%g0\n\t"
55 "be,pn %%icc,2f\n\t"
56 "nop\n\t"
57 "1:\n\t"
58 "sethi %%hi(0x7f800000),%2\n\t"
59 "andcc %0,%2,%%g0\n\t"
60 "bne,pt %%icc,1f\n\t"
61 "nop\n\t"
62 "or %%g0,1,%0\n\t"
63 "ba 2f\n\t" /* subnormal */
64 "nop\n\t"
65 "1:\n\t"
66 "subcc %0,%2,%%g0\n\t"
67 "bge,pn %%icc,1f\n\t"
68 "nop\n\t"
69 "or %%g0,2,%0\n\t"
70 "ba 2f\n\t" /* normal */
71 "nop\n\t"
72 "1:\n\t"
73 "bg,pn %%icc,1f\n\t"
74 "nop\n\t"
75 "or %%g0,3,%0\n\t"
76 "ba 2f\n\t" /* infinity */
77 "nop\n\t"
78 "1:\n\t"
79 "sethi %%hi(0x00400000),%2\n\t"
80 "andcc %0,%2,%%g0\n\t"
81 "or %%g0,4,%0\n\t"
82 "bne,pt %%icc,2f\n\t" /* quiet NaN */
83 "nop\n\t"
84 "or %%g0,5,%0\n\t" /* signalling NaN */
85 "2:\n\t"
86 : "=r" (ret), "=m" (fint), "=r" (tmp), "+f" (f)
87 :
88 : "cc");
89
90 return (ret);
91 }
92
93 extern __GNU_INLINE enum fp_class_type
fp_class(double d)94 fp_class(double d)
95 {
96 enum fp_class_type ret;
97 uint64_t dint; /* Scratch for d-as-long */
98 uint64_t tmp;
99
100 __asm__ __volatile__(
101 "fabsd %3,%3\n\t"
102 "std %3,%1\n\t"
103 "ldx %1,%0\n\t"
104 "orcc %%g0,%0,%%g0\n\t"
105 "be,pn %%xcc,2f\n\t"
106 "nop\n\t"
107 "sethi %%hi(0x7ff00000),%2\n\t"
108 "sllx %2,32,%2\n\t"
109 "andcc %0,%2,%%g0\n\t"
110 "bne,pt %%xcc,1f\n\t"
111 "nop\n\t"
112 "or %%g0,1,%0\n\t"
113 "ba 2f\n\t"
114 "nop\n\t"
115 "1:\n\t"
116 "subcc %0,%2,%%g0\n\t"
117 "bge,pn %%xcc,1f\n\t"
118 "nop\n\t"
119 "or %%g0,2,%0\n\t"
120 "ba 2f\n\t"
121 "nop\n\t"
122 "1:\n\t"
123 "andncc %0,%2,%0\n\t"
124 "bne,pn %%xcc,1f\n\t"
125 "nop\n\t"
126 "or %%g0,3,%0\n\t"
127 "ba 2f\n\t"
128 "nop\n\t"
129 "1:\n\t"
130 "sethi %%hi(0x00080000),%2\n\t"
131 "sllx %2,32,%2\n\t"
132 "andcc %0,%2,%%g0\n\t"
133 "or %%g0,4,%0\n\t"
134 "bne,pt %%xcc,2f\n\t"
135 "nop\n\t"
136 "or %%g0,5,%0\n\t"
137 "2:\n\t"
138 : "=r" (ret), "=m" (dint), "=r" (tmp), "+e" (d)
139 :
140 : "cc");
141
142 return (ret);
143 }
144
145 extern __GNU_INLINE float
__inline_sqrtf(float f)146 __inline_sqrtf(float f)
147 {
148 float ret;
149
150 __asm__ __volatile__("fsqrts %1,%0\n\t" : "=f" (ret) : "f" (f));
151 return (ret);
152 }
153
154 extern __GNU_INLINE double
__inline_sqrt(double d)155 __inline_sqrt(double d)
156 {
157 double ret;
158
159 __asm__ __volatile__("fsqrtd %1,%0\n\t" : "=f" (ret) : "f" (d));
160 return (ret);
161 }
162
163 extern __GNU_INLINE int
__swapEX(int i)164 __swapEX(int i)
165 {
166 int ret;
167 uint32_t fsr;
168 uint64_t tmp1, tmp2;
169
170 __asm__ __volatile__(
171 "and %4,0x1f,%2\n\t"
172 "sll %2,5,%2\n\t" /* shift input to aexc bit location */
173 ".volatile\n\t"
174 "st %%fsr,%1\n\t"
175 "ld %1,%0\n\t" /* %0 = fsr */
176 "andn %0,0x3e0,%3\n\t"
177 "or %2,%3,%2\n\t" /* %2 = new fsr */
178 "st %2,%1\n\t"
179 "ld %1,%%fsr\n\t"
180 "srl %0,5,%0\n\t"
181 "and %0,0x1f,%0\n\t"
182 ".nonvolatile\n\t"
183 : "=r" (ret), "=m" (fsr), "=r" (tmp1), "=r" (tmp2)
184 : "r" (i)
185 : "cc");
186
187 return (ret);
188 }
189
190 /*
191 * On the SPARC, __swapRP is a no-op; always return 0 for backward
192 * compatibility
193 */
194 /* ARGSUSED */
195 extern __GNU_INLINE enum fp_precision_type
__swapRP(enum fp_precision_type i)196 __swapRP(enum fp_precision_type i)
197 {
198 return (0);
199 }
200
201 extern __GNU_INLINE enum fp_direction_type
__swapRD(enum fp_direction_type d)202 __swapRD(enum fp_direction_type d)
203 {
204 enum fp_direction_type ret;
205 uint32_t fsr;
206 uint64_t tmp1, tmp2, tmp3;
207
208 __asm__ __volatile__(
209 "and %5,0x3,%0\n\t"
210 "sll %0,30,%2\n\t" /* shift input to RD bit location */
211 ".volatile\n\t"
212 "st %%fsr,%1\n\t"
213 "ld %1,%0\n\t" /* %0 = fsr */
214 /* mask of rounding direction bits */
215 "sethi %%hi(0xc0000000),%4\n\t"
216 "andn %0,%4,%3\n\t"
217 "or %2,%3,%2\n\t" /* %2 = new fsr */
218 "st %2,%1\n\t"
219 "ld %1,%%fsr\n\t"
220 "srl %0,30,%0\n\t"
221 "and %0,0x3,%0\n\t"
222 ".nonvolatile\n\t"
223 : "=r" (ret), "=m" (fsr), "=r" (tmp1), "=r" (tmp2), "=r" (tmp3)
224 : "r" (d)
225 : "cc");
226
227 return (ret);
228 }
229
230 extern __GNU_INLINE int
__swapTE(int i)231 __swapTE(int i)
232 {
233 int ret;
234 uint32_t fsr;
235 uint64_t tmp1, tmp2, tmp3;
236
237 __asm__ __volatile__(
238 "and %5,0x1f,%0\n\t"
239 "sll %0,23,%2\n\t" /* shift input to TEM bit location */
240 ".volatile\n\t"
241 "st %%fsr,%1\n\t"
242 "ld %1,%0\n\t" /* %0 = fsr */
243 /* mask of TEM (Trap Enable Mode bits) */
244 "sethi %%hi(0x0f800000),%4\n\t"
245 "andn %0,%4,%3\n\t"
246 "or %2,%3,%2\n\t" /* %2 = new fsr */
247 "st %2,%1\n\t"
248 "ld %1,%%fsr\n\t"
249 "srl %0,23,%0\n\t"
250 "and %0,0x1f,%0\n\t"
251 ".nonvolatile\n\t"
252 : "=r" (ret), "=m" (fsr), "=r" (tmp1), "=r" (tmp2), "=r" (tmp3)
253 : "r" (i)
254 : "cc");
255
256 return (ret);
257 }
258
259
260 extern __GNU_INLINE double
sqrt(double d)261 sqrt(double d)
262 {
263 return (__inline_sqrt(d));
264 }
265
266 extern __GNU_INLINE float
sqrtf(float f)267 sqrtf(float f)
268 {
269 return (__inline_sqrtf(f));
270 }
271
272 extern __GNU_INLINE double
fabs(double d)273 fabs(double d)
274 {
275 double ret;
276
277 __asm__ __volatile__("fabsd %1,%0\n\t" : "=e" (ret) : "e" (d));
278 return (ret);
279 }
280
281 extern __GNU_INLINE float
fabsf(float f)282 fabsf(float f)
283 {
284 float ret;
285
286 __asm__ __volatile__("fabss %1,%0\n\t" : "=f" (ret) : "f" (f));
287 return (ret);
288 }
289
290 #ifdef __cplusplus
291 }
292 #endif
293
294 #endif /* __GNUC__ */
295
296 #endif /* _LIBM_INLINES_H */
297