1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
|
/*
* FFT/MDCT transform with 3DNow! optimizations
* Copyright (c) 2006 Zuxy MENG Jie.
* Based on fft_sse.c copyright (c) 2002 Fabrice Bellard.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "../dsputil.h"
#include <math.h>
#ifdef HAVE_MM3DNOW
#include <mm3dnow.h>
static const int p1m1[2] __attribute__((aligned(8))) =
{ 0, 1 << 31 };
static const int m1p1[2] __attribute__((aligned(8))) =
{ 1 << 31, 0 };
void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z)
{
int ln = s->nbits;
int j, np, np2;
int nblocks, nloops;
register FFTComplex *p, *q;
FFTComplex *cptr, *cptr1;
int k;
np = 1 << ln;
/* FEMMS not a must here but recommended by AMD */
_m_femms();
{
__m64 *r, a0, a1, b0, b1, tmp, c;
r = (__m64 *)&z[0];
if (s->inverse)
c = *(__m64 *)m1p1;
else
c = *(__m64 *)p1m1;
j = (np >> 2);
do {
/* do the pass 0 butterfly */
a0 = _m_pfadd(r[0], r[1]);
a1 = _m_pfsub(r[0], r[1]);
/* do the pass 0 butterfly */
b0 = _m_pfadd(r[2], r[3]);
b1 = _m_pfsub(r[2], r[3]);
/* multiply third by -i */
tmp = _m_punpckhdq(b1, b1);
b1 = _m_punpckldq(b1, b1);
b1 = _m_punpckldq(tmp, b1);
b1 = _m_pxor(b1, c);
/* do the pass 1 butterfly */
r[0] = _m_pfadd(a0, b0);
r[1] = _m_pfadd(a1, b1);
r[2] = _m_pfsub(a0, b0);
r[3] = _m_pfsub(a1, b1);
r += 4;
} while (--j != 0);
}
/* pass 2 .. ln-1 */
nblocks = np >> 3;
nloops = 1 << 2;
np2 = np >> 1;
cptr1 = s->exptab1;
do {
p = z;
q = z + nloops;
j = nblocks;
do {
cptr = cptr1;
k = nloops >> 1;
do {
__m64 a0, a1, b0, b1, c0, c1, t10, t11, t20, t21;
a0 = *(__m64 *)&p[0];
a1 = *(__m64 *)&p[1];
b0 = *(__m64 *)&q[0];
b1 = *(__m64 *)&q[1];
/* complex mul */
c0 = *(__m64 *)&cptr[0];
c1 = *(__m64 *)&cptr[1];
/* cre*re cim*re */
t10 = _m_pfmul(c0, _m_punpckldq(b0, b0));
t11 = _m_pfmul(c1, _m_punpckldq(b1, b1));
c0 = *(__m64 *)&cptr[2];
c1 = *(__m64 *)&cptr[3];
/* -cim*im cre*im */
t20 = _m_pfmul(c0, _m_punpckhdq(b0, b0));
t21 = _m_pfmul(c1, _m_punpckhdq(b1, b1));
b0 = _m_pfadd(t10, t20);
b1 = _m_pfadd(t11, t21);
/* butterfly */
*(__m64 *)&p[0] = _m_pfadd(a0, b0);
*(__m64 *)&p[1] = _m_pfadd(a1, b1);
*(__m64 *)&q[0] = _m_pfsub(a0, b0);
*(__m64 *)&q[1] = _m_pfsub(a1, b1);
p += 2;
q += 2;
cptr += 4;
} while (--k);
p += nloops;
q += nloops;
} while (--j);
cptr1 += nloops * 2;
nblocks = nblocks >> 1;
nloops = nloops << 1;
} while (nblocks != 0);
_m_femms();
}
#endif
|
