1 |
148 |
jeremybenn |
/*
|
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 |
|
|
|
34 |
|
|
#ifndef _CBRT_H_
|
35 |
|
|
#define _CBRT_H_ 1
|
36 |
|
|
|
37 |
|
|
#include <spu_intrinsics.h>
|
38 |
|
|
#include "headers/vec_literal.h"
|
39 |
|
|
|
40 |
|
|
static double cbrt_factors[5] = {
|
41 |
|
|
0.629960524947436484311, /* 2^(-2/3) */
|
42 |
|
|
0.793700525984099680699, /* 2^(-1/3) */
|
43 |
|
|
1.0, /* 2^(0) */
|
44 |
|
|
1.259921049894873164666, /* 2^(1/3) */
|
45 |
|
|
1.587401051968199583441 /* 2^(2/3) */
|
46 |
|
|
};
|
47 |
|
|
|
48 |
|
|
/* Compute the cube root of x to double precision.
|
49 |
|
|
*/
|
50 |
|
|
|
51 |
|
|
static __inline double _cbrt(double x)
|
52 |
|
|
{
|
53 |
|
|
vec_int4 exp, bias;
|
54 |
|
|
vec_uint4 e_div_3, e_mod_3;
|
55 |
|
|
vec_float4 bf, inv_bf;
|
56 |
|
|
vec_float4 onef = VEC_SPLAT_F32(1.0f);
|
57 |
|
|
vec_ullong2 mask;
|
58 |
|
|
vec_ullong2 mant_mask = VEC_SPLAT_U64(0xFFFFFFFFFFFFFULL);
|
59 |
|
|
vec_double2 one = VEC_SPLAT_F64(1.0);
|
60 |
|
|
vec_double2 two = VEC_SPLAT_F64(2.0);
|
61 |
|
|
vec_double2 half = VEC_SPLAT_F64(0.5);
|
62 |
|
|
/* Polynomial coefficients */
|
63 |
|
|
vec_double2 c0 = VEC_SPLAT_F64(0.354895765043919860);
|
64 |
|
|
vec_double2 c1 = VEC_SPLAT_F64(1.50819193781584896);
|
65 |
|
|
vec_double2 c2 = VEC_SPLAT_F64(-2.11499494167371287);
|
66 |
|
|
vec_double2 c3 = VEC_SPLAT_F64(2.44693122563534430);
|
67 |
|
|
vec_double2 c4 = VEC_SPLAT_F64(-1.83469277483613086);
|
68 |
|
|
vec_double2 c5 = VEC_SPLAT_F64(0.784932344976639262);
|
69 |
|
|
vec_double2 c6 = VEC_SPLAT_F64(0.145263899385486377);
|
70 |
|
|
vec_double2 in, out, mant, u, u3, ym, a, b, factor, inv_b;
|
71 |
|
|
|
72 |
|
|
in = spu_promote(x, 0);
|
73 |
|
|
|
74 |
|
|
/* Normalize the mantissa (fraction part) into the range [0.5, 1.0) and
|
75 |
|
|
* extract the exponent.
|
76 |
|
|
*/
|
77 |
|
|
mant = spu_sel(half, in, mant_mask);
|
78 |
|
|
exp = spu_and(spu_rlmask((vec_int4)in, -20), 0x7FF);
|
79 |
|
|
|
80 |
|
|
/* Generate mask used to zero result if the exponent is zero (ie, <in> is
|
81 |
|
|
* either zero or a denorm
|
82 |
|
|
*/
|
83 |
|
|
mask = (vec_ullong2)spu_cmpeq(exp, 0);
|
84 |
|
|
mask = spu_shuffle(mask, mask, VEC_LITERAL(vec_uchar16, 0,1,2,3,0,1,2,3,8,9,10,11,8,9,10,11));
|
85 |
|
|
exp = spu_add(exp, -1022);
|
86 |
|
|
|
87 |
|
|
u = spu_madd(mant, spu_madd(mant, spu_madd(mant, spu_madd(mant, spu_madd(mant, spu_nmsub(mant, c6, c5), c4), c3), c2), c1), c0);
|
88 |
|
|
u3 = spu_mul(spu_mul(u, u), u);
|
89 |
|
|
|
90 |
|
|
/* Compute: e_div_3 = exp/3
|
91 |
|
|
*
|
92 |
|
|
* Fetch: factor = factor[2+exp%3]
|
93 |
|
|
*
|
94 |
|
|
* The factors array contains 5 values: 2^(-2/3), 2^(-1/3), 2^0, 2^(1/3),
|
95 |
|
|
* 2^(2/3), 2^1.
|
96 |
|
|
* The fetch is done using shuffle bytes so that is can easily be extended
|
97 |
|
|
* to support SIMD compution.
|
98 |
|
|
*/
|
99 |
|
|
bias = spu_rlmask(spu_rlmaska(exp, -15), -16);
|
100 |
|
|
e_div_3 = (vec_uint4)spu_rlmaska(spu_madd((vec_short8)exp, VEC_SPLAT_S16(0x5556), bias), -16);
|
101 |
|
|
|
102 |
|
|
e_mod_3 = (vec_uint4)spu_sub((vec_int4)(exp), spu_mulo((vec_short8)e_div_3, VEC_SPLAT_S16(3)));
|
103 |
|
|
|
104 |
|
|
factor = spu_promote(cbrt_factors[2+spu_extract(e_mod_3, 0)], 0);
|
105 |
|
|
|
106 |
|
|
/* Compute the estimated mantissa cube root (ym) equals:
|
107 |
|
|
* ym = (u * factor * (2.0 * mant + u3)) / (2.0 * u3 + mant);
|
108 |
|
|
*/
|
109 |
|
|
a = spu_mul(spu_mul(factor, u), spu_madd(two, mant, u3));
|
110 |
|
|
b = spu_madd(two, u3, mant);
|
111 |
|
|
|
112 |
|
|
bf = spu_roundtf(b);
|
113 |
|
|
inv_bf = spu_re(bf);
|
114 |
|
|
inv_bf = spu_madd(spu_nmsub(bf, inv_bf, onef), inv_bf, inv_bf);
|
115 |
|
|
|
116 |
|
|
inv_b = spu_extend(inv_bf);
|
117 |
|
|
inv_b = spu_madd(spu_nmsub(b, inv_b, one), inv_b, inv_b);
|
118 |
|
|
|
119 |
|
|
ym = spu_mul(a, inv_b);
|
120 |
|
|
ym = spu_madd(spu_nmsub(b, ym, a), inv_b, ym);
|
121 |
|
|
|
122 |
|
|
/* Merge sign, computed exponent, and computed mantissa.
|
123 |
|
|
*/
|
124 |
|
|
exp = spu_rl(spu_add((vec_int4)e_div_3, 1023), 20);
|
125 |
|
|
exp = spu_andc(exp, (vec_int4)mant_mask);
|
126 |
|
|
out = spu_sel((vec_double2)exp, in, VEC_SPLAT_U64(0x8000000000000000ULL));
|
127 |
|
|
out = spu_mul(out, ym);
|
128 |
|
|
|
129 |
|
|
out = spu_andc(out, (vec_double2)mask);
|
130 |
|
|
|
131 |
|
|
return (spu_extract(out, 0));
|
132 |
|
|
}
|
133 |
|
|
|
134 |
|
|
#endif /* _CBRT_H_ */
|