| 1 |
207 |
jeremybenn |
/* -------------------------------------------------------------- */
|
| 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 _ERFD2_H_
|
| 40 |
|
|
#define _ERFD2_H_ 1
|
| 41 |
|
|
|
| 42 |
|
|
#include <spu_intrinsics.h>
|
| 43 |
|
|
|
| 44 |
|
|
#include "expd2.h"
|
| 45 |
|
|
#include "recipd2.h"
|
| 46 |
|
|
#include "divd2.h"
|
| 47 |
|
|
#include "erf_utils.h"
|
| 48 |
|
|
|
| 49 |
|
|
/*
|
| 50 |
|
|
* FUNCTION
|
| 51 |
|
|
* vector double _erfd2(vector double x)
|
| 52 |
|
|
*
|
| 53 |
|
|
* DESCRIPTION
|
| 54 |
|
|
* The erfd2 function computes the error function of each element of x.
|
| 55 |
|
|
*
|
| 56 |
|
|
* C99 Special Cases:
|
| 57 |
|
|
* - erf(+0) returns +0
|
| 58 |
|
|
* - erf(-0) returns -0
|
| 59 |
|
|
* - erf(+infinite) returns +1
|
| 60 |
|
|
* - erf(-infinite) returns -1
|
| 61 |
|
|
*
|
| 62 |
|
|
* Other Cases:
|
| 63 |
|
|
* - erf(Nan) returns Nan
|
| 64 |
|
|
*
|
| 65 |
|
|
*/
|
| 66 |
|
|
|
| 67 |
|
|
static __inline vector double _erfd2(vector double x)
|
| 68 |
|
|
{
|
| 69 |
|
|
vec_uchar16 dup_even = ((vec_uchar16) { 0,1,2,3, 0,1,2,3, 8, 9,10,11, 8, 9,10,11 });
|
| 70 |
|
|
vec_double2 onehalfd = spu_splats(0.5);
|
| 71 |
|
|
vec_double2 oned = spu_splats(1.0);
|
| 72 |
|
|
vec_double2 sign_mask = spu_splats(-0.0);
|
| 73 |
|
|
|
| 74 |
|
|
/* This is where we switch from Taylor Series to Continued Fraction approximation */
|
| 75 |
|
|
vec_float4 approx_point = spu_splats(1.77f);
|
| 76 |
|
|
|
| 77 |
|
|
vec_double2 xabs, xsqu, xsign;
|
| 78 |
|
|
vec_double2 tresult, presult, result;
|
| 79 |
|
|
|
| 80 |
|
|
xsign = spu_and(x, sign_mask);
|
| 81 |
|
|
xabs = spu_andc(x, sign_mask);
|
| 82 |
|
|
xsqu = spu_mul(x, x);
|
| 83 |
|
|
|
| 84 |
|
|
/*
|
| 85 |
|
|
* Taylor Series Expansion near Zero
|
| 86 |
|
|
*/
|
| 87 |
|
|
TAYLOR_ERF(xabs, xsqu, tresult);
|
| 88 |
|
|
|
| 89 |
|
|
/*
|
| 90 |
|
|
* Continued Fraction Approximation of Erfc().
|
| 91 |
|
|
* erf = 1 - erfc
|
| 92 |
|
|
*/
|
| 93 |
|
|
CONTFRAC_ERFC(xabs, xsqu, presult);
|
| 94 |
|
|
presult = spu_sub(oned, presult);
|
| 95 |
|
|
|
| 96 |
|
|
|
| 97 |
|
|
/*
|
| 98 |
|
|
* Select the appropriate approximation.
|
| 99 |
|
|
*/
|
| 100 |
|
|
vec_float4 xf = spu_roundtf(xabs);
|
| 101 |
|
|
xf = spu_shuffle(xf, xf, dup_even);
|
| 102 |
|
|
result = spu_sel(tresult, presult, (vec_ullong2)spu_cmpgt(xf, approx_point));
|
| 103 |
|
|
|
| 104 |
|
|
|
| 105 |
|
|
/*
|
| 106 |
|
|
* Special cases/errors.
|
| 107 |
|
|
*/
|
| 108 |
|
|
|
| 109 |
|
|
/* x = +/- infinite, return +/-1 */
|
| 110 |
|
|
/* x = nan, return x */
|
| 111 |
|
|
result = spu_sel(result, oned, spu_testsv(x, SPU_SV_NEG_INFINITY | SPU_SV_POS_INFINITY));
|
| 112 |
|
|
result = spu_sel(result, x, spu_testsv(x, SPU_SV_NEG_DENORM | SPU_SV_POS_DENORM));
|
| 113 |
|
|
|
| 114 |
|
|
/*
|
| 115 |
|
|
* Preserve sign in result, since erf(-x) = -erf(x)
|
| 116 |
|
|
*/
|
| 117 |
|
|
result = spu_or(result, xsign);
|
| 118 |
|
|
|
| 119 |
|
|
return result;
|
| 120 |
|
|
}
|
| 121 |
|
|
|
| 122 |
|
|
#endif /* _ERFD2_H_ */
|
| 123 |
|
|
#endif /* __SPU__ */
|
