1 /* -------------------------------------------------------------- */
2 /* (C)Copyright 2007,2008, */
3 /* International Business Machines Corporation */
4 /* All Rights Reserved. */
5 /* */
6 /* Redistribution and use in source and binary forms, with or */
7 /* without modification, are permitted provided that the */
8 /* following conditions are met: */
9 /* */
10 /* - Redistributions of source code must retain the above copyright*/
11 /* notice, this list of conditions and the following disclaimer. */
12 /* */
13 /* - Redistributions in binary form must reproduce the above */
14 /* copyright notice, this list of conditions and the following */
15 /* disclaimer in the documentation and/or other materials */
16 /* provided with the distribution. */
17 /* */
18 /* - Neither the name of IBM Corporation nor the names of its */
19 /* contributors may be used to endorse or promote products */
20 /* derived from this software without specific prior written */
21 /* permission. */
22 /* */
23 /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */
24 /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */
25 /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
26 /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
27 /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR */
28 /* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, */
29 /* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT */
30 /* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; */
31 /* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) */
32 /* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN */
33 /* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR */
34 /* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, */
35 /* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */
36 /* -------------------------------------------------------------- */
37 /* PROLOG END TAG zYx */
38
39 #ifdef __SPU__
40 #ifndef _LOG1PD2_H_
41 #define _LOG1PD2_H_ 1
42
43 #include <spu_intrinsics.h>
44 #include "simdmath.h"
45
46 #include "logd2.h"
47 #include "divd2.h"
48
49
50
51 #define LOG1PD2_P0 0.0000000000000000000000000e+00
52 #define LOG1PD2_P1 1.0000000000000000000000000e+00
53 #define LOG1PD2_P2 2.3771612265431403265836252e+00
54 #define LOG1PD2_P3 2.0034423569559494104908026e+00
55 #define LOG1PD2_P4 7.1309327316770110272159400e-01
56 #define LOG1PD2_P5 9.8219761968547217301228613e-02
57 #define LOG1PD2_P6 3.4385125174546914139650511e-03
58
59 #define LOG1PD2_Q0 1.0000000000000000000000000e+00
60 #define LOG1PD2_Q1 2.8771612265431403265836252e+00
61 #define LOG1PD2_Q2 3.1086896368941925317130881e+00
62 #define LOG1PD2_Q3 1.5583843494335058998956356e+00
63 #define LOG1PD2_Q4 3.6047236436186669283898709e-01
64 #define LOG1PD2_Q5 3.2620075387969869884496887e-02
65 #define LOG1PD2_Q6 6.8047193336239690346356479e-04
66
67
68 /*
69 * FUNCTION
70 * vector double _log1pd2(vector double x)
71 *
72 * DESCRIPTION
73 * The function _log1pd2 computes the natural logarithm of x + 1
74 * for each of the double word elements of x.
75 *
76 */
77
_log1pd2(vector double x)78 static __inline vector double _log1pd2(vector double x)
79 {
80 vector double oned = spu_splats(1.0);
81 vector double rangehi = spu_splats(0.35);
82 vector double rangelo = spu_splats(0.0);
83 vector unsigned long long use_log;
84 vector double pr, qr;
85 vector double eresult;
86 vector double rresult;
87 vector double result;
88
89 /* Compiler Bug. Replace xbug with x when spu_cmp*() doesn't
90 * modify it's arguments! */
91 volatile vector double xbug = x;
92 use_log = spu_or(spu_cmpgt(xbug, rangehi), spu_cmpgt(rangelo, xbug));
93
94 /*
95 * Calculate directly using log(x+1)
96 */
97 eresult = _logd2(spu_add(x, oned));
98
99 /*
100 * For x in [0.0,0.35], use a rational approximation.
101 */
102 pr = spu_madd(x, spu_splats(LOG1PD2_P6), spu_splats(LOG1PD2_P5));
103 qr = spu_madd(x, spu_splats(LOG1PD2_Q6), spu_splats(LOG1PD2_Q5));
104 pr = spu_madd(pr, x, spu_splats(LOG1PD2_P4));
105 qr = spu_madd(qr, x, spu_splats(LOG1PD2_Q4));
106 pr = spu_madd(pr, x, spu_splats(LOG1PD2_P3));
107 qr = spu_madd(qr, x, spu_splats(LOG1PD2_Q3));
108 pr = spu_madd(pr, x, spu_splats(LOG1PD2_P2));
109 qr = spu_madd(qr, x, spu_splats(LOG1PD2_Q2));
110 pr = spu_madd(pr, x, spu_splats(LOG1PD2_P1));
111 qr = spu_madd(qr, x, spu_splats(LOG1PD2_Q1));
112 pr = spu_madd(pr, x, spu_splats(LOG1PD2_P0));
113 qr = spu_madd(qr, x, spu_splats(LOG1PD2_Q0));
114 rresult = _divd2(pr, qr);
115
116 /*
117 * Select either direct calculation or rational approximation.
118 */
119 result = spu_sel(rresult, eresult, use_log);
120
121 return result;
122 }
123
124 #endif /* _LOG1PD2_H_ */
125 #endif /* __SPU__ */
126
