summaryrefslogtreecommitdiffstats
path: root/libavcodec/x86/fft_3dn2.c
blob: 2abb8cfbd7e3f3665fd9fcbbe5695c8c81de6471 (plain)
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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
/*
 * FFT/MDCT transform with Extended 3DNow! optimizations
 * Copyright (c) 2006-2008 Zuxy MENG Jie, Loren Merritt
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg 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 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg 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 FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavutil/x86_cpu.h"
#include "libavcodec/dsputil.h"
#include "fft.h"

DECLARE_ALIGNED(8, static const int, m1m1)[2] = { 1<<31, 1<<31 };

#ifdef EMULATE_3DNOWEXT
#define PSWAPD(s,d)\
    "movq "#s","#d"\n"\
    "psrlq $32,"#d"\n"\
    "punpckldq "#s","#d"\n"
#define ff_fft_calc_3dn2 ff_fft_calc_3dn
#define ff_fft_dispatch_3dn2 ff_fft_dispatch_3dn
#define ff_fft_dispatch_interleave_3dn2 ff_fft_dispatch_interleave_3dn
#define ff_imdct_calc_3dn2 ff_imdct_calc_3dn
#define ff_imdct_half_3dn2 ff_imdct_half_3dn
#else
#define PSWAPD(s,d) "pswapd "#s","#d"\n"
#endif

void ff_fft_dispatch_3dn2(FFTComplex *z, int nbits);
void ff_fft_dispatch_interleave_3dn2(FFTComplex *z, int nbits);

void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z)
{
    int n = 1<<s->nbits;
    int i;
    ff_fft_dispatch_interleave_3dn2(z, s->nbits);
    __asm__ volatile("femms");
    if(n <= 8)
        for(i=0; i<n; i+=2)
            FFSWAP(FFTSample, z[i].im, z[i+1].re);
}

void ff_imdct_half_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input)
{
    x86_reg j, k;
    long n = s->mdct_size;
    long n2 = n >> 1;
    long n4 = n >> 2;
    long n8 = n >> 3;
    const uint16_t *revtab = s->revtab;
    const FFTSample *tcos = s->tcos;
    const FFTSample *tsin = s->tsin;
    const FFTSample *in1, *in2;
    FFTComplex *z = (FFTComplex *)output;

    /* pre rotation */
    in1 = input;
    in2 = input + n2 - 1;
#ifdef EMULATE_3DNOWEXT
    __asm__ volatile("movd %0, %%mm7" ::"r"(1<<31));
#endif
    for(k = 0; k < n4; k++) {
        // FIXME a single block is faster, but gcc 2.95 and 3.4.x on 32bit can't compile it
        __asm__ volatile(
            "movd         %0, %%mm0 \n"
            "movd         %2, %%mm1 \n"
            "punpckldq    %1, %%mm0 \n"
            "punpckldq    %3, %%mm1 \n"
            "movq      %%mm0, %%mm2 \n"
            PSWAPD(    %%mm1, %%mm3 )
            "pfmul     %%mm1, %%mm0 \n"
            "pfmul     %%mm3, %%mm2 \n"
#ifdef EMULATE_3DNOWEXT
            "movq      %%mm0, %%mm1 \n"
            "punpckhdq %%mm2, %%mm0 \n"
            "punpckldq %%mm2, %%mm1 \n"
            "pxor      %%mm7, %%mm0 \n"
            "pfadd     %%mm1, %%mm0 \n"
#else
            "pfpnacc   %%mm2, %%mm0 \n"
#endif
            ::"m"(in2[-2*k]), "m"(in1[2*k]),
              "m"(tcos[k]), "m"(tsin[k])
        );
        __asm__ volatile(
            "movq    %%mm0, %0    \n\t"
            :"=m"(z[revtab[k]])
        );
    }

    ff_fft_dispatch_3dn2(z, s->nbits);

#define CMUL(j,mm0,mm1)\
        "movq  (%2,"#j",2), %%mm6 \n"\
        "movq 8(%2,"#j",2), "#mm0"\n"\
        "movq        %%mm6, "#mm1"\n"\
        "movq        "#mm0",%%mm7 \n"\
        "pfmul   (%3,"#j"), %%mm6 \n"\
        "pfmul   (%4,"#j"), "#mm0"\n"\
        "pfmul   (%4,"#j"), "#mm1"\n"\
        "pfmul   (%3,"#j"), %%mm7 \n"\
        "pfsub       %%mm6, "#mm0"\n"\
        "pfadd       %%mm7, "#mm1"\n"

    /* post rotation */
    j = -n2;
    k = n2-8;
    __asm__ volatile(
        "1: \n"
        CMUL(%0, %%mm0, %%mm1)
        CMUL(%1, %%mm2, %%mm3)
        "movd   %%mm0,  (%2,%0,2) \n"
        "movd   %%mm1,12(%2,%1,2) \n"
        "movd   %%mm2,  (%2,%1,2) \n"
        "movd   %%mm3,12(%2,%0,2) \n"
        "psrlq  $32,   %%mm0 \n"
        "psrlq  $32,   %%mm1 \n"
        "psrlq  $32,   %%mm2 \n"
        "psrlq  $32,   %%mm3 \n"
        "movd   %%mm0, 8(%2,%0,2) \n"
        "movd   %%mm1, 4(%2,%1,2) \n"
        "movd   %%mm2, 8(%2,%1,2) \n"
        "movd   %%mm3, 4(%2,%0,2) \n"
        "sub $8, %1 \n"
        "add $8, %0 \n"
        "jl 1b \n"
        :"+r"(j), "+r"(k)
        :"r"(z+n8), "r"(tcos+n8), "r"(tsin+n8)
        :"memory"
    );
    __asm__ volatile("femms");
}

void ff_imdct_calc_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input)
{
    x86_reg j, k;
    long n = s->mdct_size;
    long n4 = n >> 2;

    ff_imdct_half_3dn2(s, output+n4, input);

    j = -n;
    k = n-8;
    __asm__ volatile(
        "movq %4, %%mm7 \n"
        "1: \n"
        PSWAPD((%2,%1), %%mm0)
        PSWAPD((%3,%0), %%mm1)
        "pxor    %%mm7, %%mm0 \n"
        "movq    %%mm1, (%3,%1) \n"
        "movq    %%mm0, (%2,%0) \n"
        "sub $8, %1 \n"
        "add $8, %0 \n"
        "jl 1b \n"
        :"+r"(j), "+r"(k)
        :"r"(output+n4), "r"(output+n4*3),
         "m"(*m1m1)
    );
    __asm__ volatile("femms");
}

OpenPOWER on IntegriCloud