1 /*
2   (C) Copyright 2001,2006,
3   International Business Machines Corporation,
4   Sony Computer Entertainment, Incorporated,
5   Toshiba Corporation,
6 
7   All rights reserved.
8 
9   Redistribution and use in source and binary forms, with or without
10   modification, are permitted provided that the following conditions are met:
11 
12     * Redistributions of source code must retain the above copyright notice,
13   this list of conditions and the following disclaimer.
14     * Redistributions in binary form must reproduce the above copyright
15   notice, this list of conditions and the following disclaimer in the
16   documentation and/or other materials provided with the distribution.
17     * Neither the names of the copyright holders nor the names of their
18   contributors may be used to endorse or promote products derived from this
19   software without specific prior written permission.
20 
21   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
22   IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
23   TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
24   PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
25   OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
26   EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
27   PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
28   PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29   LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30   NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31   SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 */
33 #ifndef _LOG2F_H_
34 #define _LOG2F_H_	1
35 
36 #include <spu_intrinsics.h>
37 #include "headers/dom_chkf_less_than.h"
38 
39 /*
40  * FUNCTION
41  *	float _log2f(float x)
42  *
43  * DESCRIPTION
44  *	_log2f computes log (base 2) of the input value x. The log2f
45  *	function is approximated as a polynomial of order 8
46  *	(C. Hastings, Jr, 1955).
47  *
48  *                   __8__
49  *		     \
50  *		      \
51  *	log2f(1+x) =   /     Ci*x^i
52  *                   /____
53  *                    i=1
54  *
55  *	for x in the range 0.0 to 1.0
56  *
57  *	C1 =  1.4426898816672
58  *	C2 = -0.72116591947498
59  *	C3 =  0.47868480909345
60  *	C4 = -0.34730547155299
61  *	C5 =  0.24187369696082
62  *	C6 = -0.13753123777116
63  *	C7 =  0.052064690894143
64  *	C8 = -0.0093104962134977
65  *
66  *	This function assumes that x is a non-zero positive value.
67  */
68 
_log2f(float x)69 static __inline float _log2f(float x)
70 {
71   union {
72     unsigned int ui;
73     float f;
74   } in;
75   int exponent;
76   float result;
77   float x2, x4;
78   float hi, lo;
79   vector float vx;
80   vector float vc = { 0.0, 0.0, 0.0, 0.0 };
81 
82   in.f = x;
83 
84   /* Extract the exponent from the input X.
85    */
86   exponent = (signed)((in.ui >> 23) & 0xFF) - 127;
87 
88   /* Compute the remainder after removing the exponent.
89    */
90   in.ui -= exponent << 23;
91 
92   /* Calculate the log2 of the remainder using the polynomial
93    * approximation.
94    */
95   x = in.f - 1.0f;
96 
97   /* Instruction counts can be reduced if the polynomial was
98    * computed entirely from nested (dependent) fma's. However,
99    * to reduce the number of pipeline stalls, the polygon is evaluated
100    * in two halves (hi amd lo).
101    */
102   x2 = x * x;
103   x4 = x2 * x2;
104   hi = -0.0093104962134977f*x + 0.052064690894143f;
105   hi =                   hi*x - 0.13753123777116f;
106   hi =                   hi*x + 0.24187369696082f;
107   hi =                   hi*x - 0.34730547155299f;
108   lo =  0.47868480909345f  *x - 0.72116591947498f;
109   lo =                   lo*x + 1.4426898816672f;
110   lo =                   lo*x;
111   result = hi*x4 + lo;
112 
113   /* Add the exponent back into the result.
114    */
115   result += (float)(exponent);
116 
117 #ifndef _IEEE_LIBM
118   vx = spu_promote(x, 0);
119   dom_chkf_less_than(vx, vc);
120 #endif
121   return (result);
122 }
123 
124 #endif /* _LOG2F_H_ */
125