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 #ifdef __SPU__
39 #ifndef _ACOSHD2_H_
40 #define _ACOSHD2_H_	1
41 
42 #include <spu_intrinsics.h>
43 #include "logd2.h"
44 #include "sqrtd2.h"
45 
46 /*
47  * FUNCTION
48  *  vector double _acoshd2(vector double x)
49  *
50  * DESCRIPTION
51  *  The acoshd2 function returns a vector containing the hyperbolic
52  *  arccosines of the corresponding elements of the input vector.
53  *
54  *  We are using the formula:
55  *    acosh = ln(x + sqrt(x^2 - 1))
56  *
57  *  For x near one, we use the Taylor series:
58  *
59  *                infinity
60  *                 ------
61  *                  -   '
62  *                   -                 k
63  *    acosh x =       -      C  (x - 1)
64  *                   -        k
65  *                  -   ,
66  *                 ------
67  *                 k = 0
68  *
69  *
70  *  Special Cases:
71  *	- acosh(1)        = +0
72  *	- acosh(NaN)      = NaN
73  *	- acosh(Infinity) = Infinity
74  *	- acosh(x < 1)    = NaN
75  *
76  */
77 
78 /*
79  * Taylor Series Coefficients
80  * for x around 1.
81  */
82 #define SDM_ACOSHD2_TAY01  1.000000000000000000000000000000000E0  /* 1 / 1                            */
83 #define SDM_ACOSHD2_TAY02 -8.333333333333333333333333333333333E-2 /* 1 / 12                           */
84 #define SDM_ACOSHD2_TAY03  1.875000000000000000000000000000000E-2 /* 3 / 160                          */
85 #define SDM_ACOSHD2_TAY04 -5.580357142857142857142857142857142E-3 /* 5 / 896                          */
86 #define SDM_ACOSHD2_TAY05  1.898871527777777777777777777777777E-3 /* 35 / 18432                       */
87 #define SDM_ACOSHD2_TAY06 -6.991299715909090909090909090909090E-4 /* 63 / 90112                       */
88 #define SDM_ACOSHD2_TAY07  2.711369441105769230769230769230769E-4 /* 231 / 851968                     */
89 #define SDM_ACOSHD2_TAY08 -1.091003417968750000000000000000000E-4 /* 143 / 1310720                    */
90 #define SDM_ACOSHD2_TAY09  4.512422225054572610294117647058823E-5 /* 6435 / 142606336                 */
91 #define SDM_ACOSHD2_TAY10 -1.906564361170718544407894736842105E-5 /* 12155 / 637534208                */
92 #define SDM_ACOSHD2_TAY11  8.193687314078921363467261904761904E-6 /* 46189 / 5637144576               */
93 #define SDM_ACOSHD2_TAY12 -3.570569274218186088230298913043478E-6 /* 88179 / 24696061952              */
94 #define SDM_ACOSHD2_TAY13  1.574025955051183700561523437500000E-6 /* 676039 / 429496729600            */
95 #define SDM_ACOSHD2_TAY14 -7.006881922414457356488263165509259E-7 /* 1300075 / 1855425871872          */
96 #define SDM_ACOSHD2_TAY15  3.145330616650332150788142763335129E-7 /* 5014575 / 15942918602752         */
97 
_acoshd2(vector double x)98 static __inline vector double _acoshd2(vector double x)
99 {
100     vec_uchar16 dup_even  = ((vec_uchar16) { 0,1,2,3,  0,1,2,3, 8,9,10,11, 8,9,10,11 });
101     vec_double2 minus_oned = spu_splats(-1.0);
102     vec_double2 twod       = spu_splats(2.0);
103     /* Where we switch from taylor to formula */
104     vec_float4  switch_approx = spu_splats(1.15f);
105     vec_double2 result, fresult, mresult;;
106 
107 
108     vec_double2 xminus1 = spu_add(x, minus_oned);
109     vec_float4  xf = spu_roundtf(x);
110     xf = spu_shuffle(xf, xf, dup_even);
111 
112     vec_ullong2 use_form = (vec_ullong2)spu_cmpgt(xf, switch_approx);
113 
114     vec_double2 sqrtargformula = spu_madd(x, x, minus_oned);
115     vec_double2 sqrtargtaylor  = spu_mul(xminus1, twod);
116     vec_double2 sqrtarg = spu_sel(sqrtargtaylor, sqrtargformula, use_form);
117 
118     vec_double2 sqrtresult = _sqrtd2(sqrtarg);
119 
120     /*
121      * Formula:
122      *   acosh = ln(x + sqrt(x^2 - 1))
123      */
124     fresult = spu_add(x, sqrtresult);
125     fresult = _logd2(fresult);
126 
127     /*
128      * Taylor Series
129      */
130     mresult = spu_splats(SDM_ACOSHD2_TAY15);
131     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY14));
132     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY13));
133     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY12));
134     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY11));
135     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY10));
136     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY09));
137     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY08));
138     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY07));
139     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY06));
140     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY05));
141     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY04));
142     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY03));
143     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY02));
144     mresult = spu_madd(xminus1, mresult, spu_splats(SDM_ACOSHD2_TAY01));
145 
146 
147     mresult = spu_mul(mresult, sqrtresult);
148 
149 
150     /*
151      * Select series or formula
152      */
153     result = spu_sel(mresult, fresult, use_form);
154 
155     return result;
156 }
157 
158 #endif /* _ACOSHD2_H_ */
159 #endif /* __SPU__ */
160