xref: /illumos-gate/usr/src/lib/libmvec/common/__vsincosf.c (revision 66582b606a8194f7f3ba5b3a3a6dca5b0d346361)
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  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
23  */
24 /*
25  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
26  * Use is subject to license terms.
27  */
28 
29 /*
30  * __vsincosf: single precision vector sincos
31  *
32  * Algorithm:
33  *
34  * For |x| < pi/4, approximate sin(x) by a polynomial x+x*z*(S0+
35  * z*(S1+z*S2)) and cos(x) by a polynomial 1+z*(-1/2+z*(C0+z*(C1+
36  * z*C2))), where z = x*x, all evaluated in double precision.
37  *
38  * Accuracy:
39  *
40  * The largest error is less than 0.6 ulps.
41  */
42 
43 #include <sys/isa_defs.h>
44 
45 #ifdef _LITTLE_ENDIAN
46 #define	HI(x)	*(1+(int *)&x)
47 #define	LO(x)	*(unsigned *)&x
48 #else
49 #define	HI(x)	*(int *)&x
50 #define	LO(x)	*(1+(unsigned *)&x)
51 #endif
52 
53 #ifdef __RESTRICT
54 #define	restrict _Restrict
55 #else
56 #define	restrict
57 #endif
58 
59 extern int __vlibm_rem_pio2m(double *, double *, int, int, int);
60 
61 static const double C[] = {
62 	-1.66666552424430847168e-01,	/* 2^ -3 * -1.5555460000000 */
63 	8.33219196647405624390e-03,	/* 2^ -7 *  1.11077E0000000 */
64 	-1.95187909412197768688e-04,	/* 2^-13 * -1.9956B60000000 */
65 	1.0,
66 	-0.5,
67 	4.16666455566883087158e-02,	/* 2^ -5 *  1.55554A0000000 */
68 	-1.38873036485165357590e-03,	/* 2^-10 * -1.6C0C1E0000000 */
69 	2.44309903791872784495e-05,	/* 2^-16 *  1.99E24E0000000 */
70 	0.636619772367581343075535,	/* 2^ -1  * 1.45F306DC9C883 */
71 	6755399441055744.0,		/* 2^ 52  * 1.8000000000000 */
72 	1.570796326734125614166,	/* 2^  0  * 1.921FB54400000 */
73 	6.077100506506192601475e-11,	/* 2^-34  * 1.0B4611A626331 */
74 };
75 
76 #define	S0	C[0]
77 #define	S1	C[1]
78 #define	S2	C[2]
79 #define	one	C[3]
80 #define	mhalf	C[4]
81 #define	C0	C[5]
82 #define	C1	C[6]
83 #define	C2	C[7]
84 #define	invpio2	C[8]
85 #define	c3two51	C[9]
86 #define	pio2_1  C[10]
87 #define	pio2_t	C[11]
88 
89 #define	PREPROCESS(N, sindex, cindex, label)				\
90 	hx = *(int *)x;							\
91 	ix = hx & 0x7fffffff;						\
92 	t = *x;								\
93 	x += stridex;							\
94 	if (ix <= 0x3f490fdb) { /* |x| < pi/4 */			\
95 		if (ix == 0) {						\
96 			s[sindex] = t;					\
97 			c[cindex] = one;				\
98 			goto label;					\
99 		}							\
100 		y##N = (double)t;					\
101 		n##N = 0;						\
102 	} else if (ix <= 0x49c90fdb) { /* |x| < 2^19*pi */		\
103 		y##N = (double)t;					\
104 		medium = 1;						\
105 	} else {							\
106 		if (ix >= 0x7f800000) { /* inf or nan */		\
107 			s[sindex] = c[cindex] = t / t;			\
108 			goto label;					\
109 		}							\
110 		z##N = y##N = (double)t;				\
111 		hx = HI(y##N);						\
112 		n##N = ((hx >> 20) & 0x7ff) - 1046;			\
113 		HI(z##N) = (hx & 0xfffff) | 0x41600000;			\
114 		n##N = __vlibm_rem_pio2m(&z##N, &y##N, n##N, 1, 0);	\
115 		if (hx < 0) {						\
116 			y##N = -y##N;					\
117 			n##N = -n##N;					\
118 		}							\
119 		z##N = y##N * y##N;					\
120 		f##N = (float)(y##N + y##N * z##N * (S0 + z##N *	\
121 		    (S1 + z##N * S2)));					\
122 		g##N = (float)(one + z##N * (mhalf + z##N * (C0 +	\
123 		    z##N * (C1 + z##N * C2))));				\
124 		if (n##N & 2) {						\
125 			f##N = -f##N;					\
126 			g##N = -g##N;					\
127 		}							\
128 		if (n##N & 1) {						\
129 			s[sindex] = g##N;				\
130 			c[cindex] = -f##N;				\
131 		} else {						\
132 			s[sindex] = f##N;				\
133 			c[cindex] = g##N;				\
134 		}							\
135 		goto label;						\
136 	}
137 
138 #define	PROCESS(N)							\
139 	if (medium) {							\
140 		z##N = y##N * invpio2 + c3two51;			\
141 		n##N = LO(z##N);					\
142 		z##N -= c3two51;					\
143 		y##N = (y##N - z##N * pio2_1) - z##N * pio2_t;		\
144 	}								\
145 	z##N = y##N * y##N;						\
146 	f##N = (float)(y##N + y##N * z##N * (S0 + z##N * (S1 + z##N * S2)));\
147 	g##N = (float)(one + z##N * (mhalf + z##N * (C0 + z##N *	\
148 	    (C1 + z##N * C2))));					\
149 	if (n##N & 2) {							\
150 		f##N = -f##N;						\
151 		g##N = -g##N;						\
152 	}								\
153 	if (n##N & 1) {							\
154 		*s = g##N;						\
155 		*c = -f##N;						\
156 	} else {							\
157 		*s = f##N;						\
158 		*c = g##N;						\
159 	}								\
160 	s += strides;							\
161 	c += stridec
162 
163 void
164 __vsincosf(int n, float *restrict x, int stridex,
165     float *restrict s, int strides, float *restrict c, int stridec)
166 {
167 	double		y0, y1, y2, y3;
168 	double		z0, z1, z2, z3;
169 	float		f0, f1, f2, f3, t;
170 	float		g0, g1, g2, g3;
171 	int		n0 = 0, n1 = 0, n2 = 0, n3, hx, ix, medium;
172 
173 	s -= strides;
174 	c -= stridec;
175 
176 	for (;;) {
177 begin:
178 		s += strides;
179 		c += stridec;
180 
181 		if (--n < 0)
182 			break;
183 
184 		medium = 0;
185 		PREPROCESS(0, 0, 0, begin);
186 
187 		if (--n < 0)
188 			goto process1;
189 
190 		PREPROCESS(1, strides, stridec, process1);
191 
192 		if (--n < 0)
193 			goto process2;
194 
195 		PREPROCESS(2, (strides << 1), (stridec << 1), process2);
196 
197 		if (--n < 0)
198 			goto process3;
199 
200 		PREPROCESS(3, (strides << 1) + strides,
201 		    (stridec << 1) + stridec, process3);
202 
203 		if (medium) {
204 			z0 = y0 * invpio2 + c3two51;
205 			z1 = y1 * invpio2 + c3two51;
206 			z2 = y2 * invpio2 + c3two51;
207 			z3 = y3 * invpio2 + c3two51;
208 
209 			n0 = LO(z0);
210 			n1 = LO(z1);
211 			n2 = LO(z2);
212 			n3 = LO(z3);
213 
214 			z0 -= c3two51;
215 			z1 -= c3two51;
216 			z2 -= c3two51;
217 			z3 -= c3two51;
218 
219 			y0 = (y0 - z0 * pio2_1) - z0 * pio2_t;
220 			y1 = (y1 - z1 * pio2_1) - z1 * pio2_t;
221 			y2 = (y2 - z2 * pio2_1) - z2 * pio2_t;
222 			y3 = (y3 - z3 * pio2_1) - z3 * pio2_t;
223 		}
224 
225 		z0 = y0 * y0;
226 		z1 = y1 * y1;
227 		z2 = y2 * y2;
228 		z3 = y3 * y3;
229 
230 		f0 = (float)(y0 + y0 * z0 * (S0 + z0 * (S1 + z0 * S2)));
231 		f1 = (float)(y1 + y1 * z1 * (S0 + z1 * (S1 + z1 * S2)));
232 		f2 = (float)(y2 + y2 * z2 * (S0 + z2 * (S1 + z2 * S2)));
233 		f3 = (float)(y3 + y3 * z3 * (S0 + z3 * (S1 + z3 * S2)));
234 
235 		g0 = (float)(one + z0 * (mhalf + z0 * (C0 + z0 *
236 		    (C1 + z0 * C2))));
237 		g1 = (float)(one + z1 * (mhalf + z1 * (C0 + z1 *
238 		    (C1 + z1 * C2))));
239 		g2 = (float)(one + z2 * (mhalf + z2 * (C0 + z2 *
240 		    (C1 + z2 * C2))));
241 		g3 = (float)(one + z3 * (mhalf + z3 * (C0 + z3 *
242 		    (C1 + z3 * C2))));
243 
244 		if (n0 & 2) {
245 			f0 = -f0;
246 			g0 = -g0;
247 		}
248 		if (n1 & 2) {
249 			f1 = -f1;
250 			g1 = -g1;
251 		}
252 		if (n2 & 2) {
253 			f2 = -f2;
254 			g2 = -g2;
255 		}
256 		if (n3 & 2) {
257 			f3 = -f3;
258 			g3 = -g3;
259 		}
260 
261 		if (n0 & 1) {
262 			*s = g0;
263 			*c = -f0;
264 		} else {
265 			*s = f0;
266 			*c = g0;
267 		}
268 		s += strides;
269 		c += stridec;
270 
271 		if (n1 & 1) {
272 			*s = g1;
273 			*c = -f1;
274 		} else {
275 			*s = f1;
276 			*c = g1;
277 		}
278 		s += strides;
279 		c += stridec;
280 
281 		if (n2 & 1) {
282 			*s = g2;
283 			*c = -f2;
284 		} else {
285 			*s = f2;
286 			*c = g2;
287 		}
288 		s += strides;
289 		c += stridec;
290 
291 		if (n3 & 1) {
292 			*s = g3;
293 			*c = -f3;
294 		} else {
295 			*s = f3;
296 			*c = g3;
297 		}
298 		continue;
299 
300 process1:
301 		PROCESS(0);
302 		continue;
303 
304 process2:
305 		PROCESS(0);
306 		PROCESS(1);
307 		continue;
308 
309 process3:
310 		PROCESS(0);
311 		PROCESS(1);
312 		PROCESS(2);
313 	}
314 }
315