← Back to branch summary

~ubuntu-branches/ubuntu/jaunty/xvidcap/jaunty-proposed

« back to all changes in this revision

Viewing changes to ffmpeg/libavcodec/i386/mpegvideo_mmx_template.c

  • Committer: Bazaar Package Importer
  • Author(s): John Dong
  • Date: 2008-02-25 15:47:12 UTC
  • mfrom: (1.1.1 upstream)
  • Revision ID: james.westby@ubuntu.com-20080225154712-qvr11ekcea4c9ry8
Tags: 1.1.6-0.1ubuntu1
https://launchpad.net/bugs/120003
* Merge from debian-multimedia (LP: #120003), Ubuntu Changes:
 - For ffmpeg-related build-deps, remove cvs from package names.
 - Standards-Version 3.7.3
 - Maintainer Spec

Show diffs side-by-side

added added

removed removed

Lines of Context:
ffmpeg/libavcodec/i386/mpegvideo_mmx_template.c
3
3
 *
4
4
 * Copyright (c) 2002 Michael Niedermayer <michaelni@gmx.at>
5
5
 *
6
 
 * This library is free software; you can redistribute it and/or
 
6
 * This file is part of FFmpeg.
 
7
 *
 
8
 * FFmpeg is free software; you can redistribute it and/or
7
9
 * modify it under the terms of the GNU Lesser General Public
8
10
 * License as published by the Free Software Foundation; either
9
 
 * version 2 of the License, or (at your option) any later version.
 
11
 * version 2.1 of the License, or (at your option) any later version.
10
12
 *
11
 
 * This library is distributed in the hope that it will be useful,
 
13
 * FFmpeg is distributed in the hope that it will be useful,
12
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13
15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
16
 * Lesser General Public License for more details.
15
17
 *
16
18
 * You should have received a copy of the GNU Lesser General Public
17
 
 * License along with this library; if not, write to the Free Software
18
 
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
19
 * License along with FFmpeg; if not, write to the Free Software
 
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
21
 */
20
22
#undef SPREADW
21
23
#undef PMAXW
 
24
#undef PMAX
22
25
#ifdef HAVE_MMX2
23
26
#define SPREADW(a) "pshufw $0, " #a ", " #a " \n\t"
24
 
#define PMAXW(a,b) "pmaxsw " #a ", " #b " \n\t"
25
 
 
 
27
#define PMAXW(a,b) "pmaxsw " #a ", " #b "     \n\t"
 
28
#define PMAX(a,b) \
 
29
            "pshufw $0x0E," #a ", " #b "        \n\t"\
 
30
            PMAXW(b, a)\
 
31
            "pshufw $0x01," #a ", " #b "        \n\t"\
 
32
            PMAXW(b, a)
26
33
#else
27
34
#define SPREADW(a) \
28
 
        "punpcklwd " #a ", " #a " \n\t"\
29
 
        "punpcklwd " #a ", " #a " \n\t"
 
35
        "punpcklwd " #a ", " #a " \n\t"\
 
36
        "punpcklwd " #a ", " #a " \n\t"
30
37
#define PMAXW(a,b) \
31
 
        "psubusw " #a ", " #b " \n\t"\
32
 
        "paddw " #a ", " #b " \n\t"
 
38
        "psubusw " #a ", " #b " \n\t"\
 
39
        "paddw " #a ", " #b "   \n\t"
 
40
#define PMAX(a,b)  \
 
41
            "movq " #a ", " #b "                \n\t"\
 
42
            "psrlq $32, " #a "                  \n\t"\
 
43
            PMAXW(b, a)\
 
44
            "movq " #a ", " #b "                \n\t"\
 
45
            "psrlq $16, " #a "                  \n\t"\
 
46
            PMAXW(b, a)
 
47
 
33
48
#endif
34
49
 
35
50
static int RENAME(dct_quantize)(MpegEncContext *s,
36
51
                            DCTELEM *block, int n,
37
52
                            int qscale, int *overflow)
38
53
{
39
 
    int level=0, last_non_zero_p1, q; //=0 is cuz gcc says uninitalized ...
 
54
    long last_non_zero_p1;
 
55
    int level=0, q; //=0 is cuz gcc says uninitalized ...
40
56
    const uint16_t *qmat, *bias;
41
 
    __align8 int16_t temp_block[64];
42
 
    
 
57
    DECLARE_ALIGNED_8(int16_t, temp_block[64]);
 
58
 
43
59
    assert((7&(int)(&temp_block[0])) == 0); //did gcc align it correctly?
44
60
 
45
61
    //s->fdct (block);
46
 
    ff_fdct_mmx (block); //cant be anything else ...
 
62
    RENAMEl(ff_fdct) (block); //cant be anything else ...
 
63
 
 
64
    if(s->dct_error_sum)
 
65
        s->denoise_dct(s, block);
47
66
 
48
67
    if (s->mb_intra) {
49
68
        int dummy;
55
74
        if (!s->h263_aic) {
56
75
#if 1
57
76
        asm volatile (
58
 
                "imul %%ecx             \n\t"
59
 
                : "=d" (level), "=a"(dummy)
60
 
                : "a" ((block[0]>>2) + q), "c" (inverse[q<<1])
 
77
                "mul %%ecx                \n\t"
 
78
                : "=d" (level), "=a"(dummy)
 
79
                : "a" ((block[0]>>2) + q), "c" (ff_inverse[q<<1])
61
80
        );
62
81
#else
63
82
        asm volatile (
64
 
                "xorl %%edx, %%edx      \n\t"
65
 
                "divw %%cx              \n\t"
66
 
                "movzwl %%ax, %%eax     \n\t"
67
 
                : "=a" (level)
68
 
                : "a" ((block[0]>>2) + q), "c" (q<<1)
69
 
                : "%edx"
 
83
                "xorl %%edx, %%edx        \n\t"
 
84
                "divw %%cx                \n\t"
 
85
                "movzwl %%ax, %%eax       \n\t"
 
86
                : "=a" (level)
 
87
                : "a" ((block[0]>>2) + q), "c" (q<<1)
 
88
                : "%edx"
70
89
        );
71
90
#endif
72
91
        } else
73
92
            /* For AIC we skip quant/dequant of INTRADC */
74
93
            level = (block[0] + 4)>>3;
75
 
            
 
94
 
76
95
        block[0]=0; //avoid fake overflow
77
96
//        temp_block[0] = (block[0] + (q >> 1)) / q;
78
97
        last_non_zero_p1 = 1;
79
 
        bias = s->q_intra_matrix16_bias[qscale];
80
 
        qmat = s->q_intra_matrix16[qscale];
 
98
        bias = s->q_intra_matrix16[qscale][1];
 
99
        qmat = s->q_intra_matrix16[qscale][0];
81
100
    } else {
82
101
        last_non_zero_p1 = 0;
83
 
        bias = s->q_inter_matrix16_bias[qscale];
84
 
        qmat = s->q_inter_matrix16[qscale];
 
102
        bias = s->q_inter_matrix16[qscale][1];
 
103
        qmat = s->q_inter_matrix16[qscale][0];
85
104
    }
86
105
 
87
 
    if(s->out_format == FMT_H263 && s->mpeg_quant==0){
88
 
    
 
106
    if((s->out_format == FMT_H263 || s->out_format == FMT_H261) && s->mpeg_quant==0){
 
107
 
89
108
        asm volatile(
90
 
            "movd %%eax, %%mm3                  \n\t" // last_non_zero_p1
 
109
            "movd %%"REG_a", %%mm3              \n\t" // last_non_zero_p1
91
110
            SPREADW(%%mm3)
92
 
            "pxor %%mm7, %%mm7                  \n\t" // 0
93
 
            "pxor %%mm4, %%mm4                  \n\t" // 0
94
 
            "movq (%2), %%mm5                   \n\t" // qmat[0]
95
 
            "pxor %%mm6, %%mm6                  \n\t"
96
 
            "psubw (%3), %%mm6                  \n\t" // -bias[0]
97
 
            "movl $-128, %%eax                  \n\t"
98
 
            ".balign 16                         \n\t"
99
 
            "1:                                 \n\t"
100
 
            "pxor %%mm1, %%mm1                  \n\t" // 0
101
 
            "movq (%1, %%eax), %%mm0            \n\t" // block[i]
102
 
            "pcmpgtw %%mm0, %%mm1               \n\t" // block[i] <= 0 ? 0xFF : 0x00
103
 
            "pxor %%mm1, %%mm0                  \n\t" 
104
 
            "psubw %%mm1, %%mm0                 \n\t" // ABS(block[i])
105
 
            "psubusw %%mm6, %%mm0               \n\t" // ABS(block[i]) + bias[0]
106
 
            "pmulhw %%mm5, %%mm0                \n\t" // (ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16
107
 
            "por %%mm0, %%mm4                   \n\t" 
108
 
            "pxor %%mm1, %%mm0                  \n\t" 
109
 
            "psubw %%mm1, %%mm0                 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
110
 
            "movq %%mm0, (%5, %%eax)            \n\t"
111
 
            "pcmpeqw %%mm7, %%mm0               \n\t" // out==0 ? 0xFF : 0x00
112
 
            "movq (%4, %%eax), %%mm1            \n\t" 
113
 
            "movq %%mm7, (%1, %%eax)            \n\t" // 0
114
 
            "pandn %%mm1, %%mm0                 \n\t"
115
 
            PMAXW(%%mm0, %%mm3)
116
 
            "addl $8, %%eax                     \n\t"
117
 
            " js 1b                             \n\t"
118
 
            "movq %%mm3, %%mm0                  \n\t"
119
 
            "psrlq $32, %%mm3                   \n\t"
120
 
            PMAXW(%%mm0, %%mm3)
121
 
            "movq %%mm3, %%mm0                  \n\t"
122
 
            "psrlq $16, %%mm3                   \n\t"
123
 
            PMAXW(%%mm0, %%mm3)
124
 
            "movd %%mm3, %%eax                  \n\t"
125
 
            "movzbl %%al, %%eax                 \n\t" // last_non_zero_p1
126
 
            : "+a" (last_non_zero_p1)
 
111
            "pxor %%mm7, %%mm7                  \n\t" // 0
 
112
            "pxor %%mm4, %%mm4                  \n\t" // 0
 
113
            "movq (%2), %%mm5                   \n\t" // qmat[0]
 
114
            "pxor %%mm6, %%mm6                  \n\t"
 
115
            "psubw (%3), %%mm6                  \n\t" // -bias[0]
 
116
            "mov $-128, %%"REG_a"               \n\t"
 
117
            ASMALIGN(4)
 
118
            "1:                                 \n\t"
 
119
            "pxor %%mm1, %%mm1                  \n\t" // 0
 
120
            "movq (%1, %%"REG_a"), %%mm0        \n\t" // block[i]
 
121
            "pcmpgtw %%mm0, %%mm1               \n\t" // block[i] <= 0 ? 0xFF : 0x00
 
122
            "pxor %%mm1, %%mm0                  \n\t"
 
123
            "psubw %%mm1, %%mm0                 \n\t" // ABS(block[i])
 
124
            "psubusw %%mm6, %%mm0               \n\t" // ABS(block[i]) + bias[0]
 
125
            "pmulhw %%mm5, %%mm0                \n\t" // (ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16
 
126
            "por %%mm0, %%mm4                   \n\t"
 
127
            "pxor %%mm1, %%mm0                  \n\t"
 
128
            "psubw %%mm1, %%mm0                 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
 
129
            "movq %%mm0, (%5, %%"REG_a")        \n\t"
 
130
            "pcmpeqw %%mm7, %%mm0               \n\t" // out==0 ? 0xFF : 0x00
 
131
            "movq (%4, %%"REG_a"), %%mm1        \n\t"
 
132
            "movq %%mm7, (%1, %%"REG_a")        \n\t" // 0
 
133
            "pandn %%mm1, %%mm0                 \n\t"
 
134
            PMAXW(%%mm0, %%mm3)
 
135
            "add $8, %%"REG_a"                  \n\t"
 
136
            " js 1b                             \n\t"
 
137
            PMAX(%%mm3, %%mm0)
 
138
            "movd %%mm3, %%"REG_a"              \n\t"
 
139
            "movzb %%al, %%"REG_a"              \n\t" // last_non_zero_p1
 
140
            : "+a" (last_non_zero_p1)
127
141
            : "r" (block+64), "r" (qmat), "r" (bias),
128
142
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64)
129
143
        );
130
144
        // note the asm is split cuz gcc doesnt like that many operands ...
131
145
        asm volatile(
132
 
            "movd %1, %%mm1                     \n\t" // max_qcoeff
133
 
            SPREADW(%%mm1)
134
 
            "psubusw %%mm1, %%mm4               \n\t" 
135
 
            "packuswb %%mm4, %%mm4              \n\t"
136
 
            "movd %%mm4, %0                     \n\t" // *overflow
 
146
            "movd %1, %%mm1                     \n\t" // max_qcoeff
 
147
            SPREADW(%%mm1)
 
148
            "psubusw %%mm1, %%mm4               \n\t"
 
149
            "packuswb %%mm4, %%mm4              \n\t"
 
150
            "movd %%mm4, %0                     \n\t" // *overflow
137
151
        : "=g" (*overflow)
138
152
        : "g" (s->max_qcoeff)
139
153
        );
140
154
    }else{ // FMT_H263
141
155
        asm volatile(
142
 
            "movd %%eax, %%mm3                  \n\t" // last_non_zero_p1
 
156
            "movd %%"REG_a", %%mm3              \n\t" // last_non_zero_p1
143
157
            SPREADW(%%mm3)
144
 
            "pxor %%mm7, %%mm7                  \n\t" // 0
145
 
            "pxor %%mm4, %%mm4                  \n\t" // 0
146
 
            "movl $-128, %%eax                  \n\t"
147
 
            ".balign 16                         \n\t"
148
 
            "1:                                 \n\t"
149
 
            "pxor %%mm1, %%mm1                  \n\t" // 0
150
 
            "movq (%1, %%eax), %%mm0            \n\t" // block[i]
151
 
            "pcmpgtw %%mm0, %%mm1               \n\t" // block[i] <= 0 ? 0xFF : 0x00
152
 
            "pxor %%mm1, %%mm0                  \n\t" 
153
 
            "psubw %%mm1, %%mm0                 \n\t" // ABS(block[i])
154
 
            "movq (%3, %%eax), %%mm6            \n\t" // bias[0]
155
 
            "paddusw %%mm6, %%mm0               \n\t" // ABS(block[i]) + bias[0]
156
 
            "movq (%2, %%eax), %%mm5            \n\t" // qmat[i]
157
 
            "pmulhw %%mm5, %%mm0                \n\t" // (ABS(block[i])*qmat[0] + bias[0]*qmat[0])>>16
158
 
            "por %%mm0, %%mm4                   \n\t" 
159
 
            "pxor %%mm1, %%mm0                  \n\t" 
160
 
            "psubw %%mm1, %%mm0                 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
161
 
            "movq %%mm0, (%5, %%eax)            \n\t"
162
 
            "pcmpeqw %%mm7, %%mm0               \n\t" // out==0 ? 0xFF : 0x00
163
 
            "movq (%4, %%eax), %%mm1            \n\t" 
164
 
            "movq %%mm7, (%1, %%eax)            \n\t" // 0
165
 
            "pandn %%mm1, %%mm0                 \n\t"
166
 
            PMAXW(%%mm0, %%mm3)
167
 
            "addl $8, %%eax                     \n\t"
168
 
            " js 1b                             \n\t"
169
 
            "movq %%mm3, %%mm0                  \n\t"
170
 
            "psrlq $32, %%mm3                   \n\t"
171
 
            PMAXW(%%mm0, %%mm3)
172
 
            "movq %%mm3, %%mm0                  \n\t"
173
 
            "psrlq $16, %%mm3                   \n\t"
174
 
            PMAXW(%%mm0, %%mm3)
175
 
            "movd %%mm3, %%eax                  \n\t"
176
 
            "movzbl %%al, %%eax                 \n\t" // last_non_zero_p1
177
 
            : "+a" (last_non_zero_p1)
 
158
            "pxor %%mm7, %%mm7                  \n\t" // 0
 
159
            "pxor %%mm4, %%mm4                  \n\t" // 0
 
160
            "mov $-128, %%"REG_a"               \n\t"
 
161
            ASMALIGN(4)
 
162
            "1:                                 \n\t"
 
163
            "pxor %%mm1, %%mm1                  \n\t" // 0
 
164
            "movq (%1, %%"REG_a"), %%mm0        \n\t" // block[i]
 
165
            "pcmpgtw %%mm0, %%mm1               \n\t" // block[i] <= 0 ? 0xFF : 0x00
 
166
            "pxor %%mm1, %%mm0                  \n\t"
 
167
            "psubw %%mm1, %%mm0                 \n\t" // ABS(block[i])
 
168
            "movq (%3, %%"REG_a"), %%mm6        \n\t" // bias[0]
 
169
            "paddusw %%mm6, %%mm0               \n\t" // ABS(block[i]) + bias[0]
 
170
            "movq (%2, %%"REG_a"), %%mm5        \n\t" // qmat[i]
 
171
            "pmulhw %%mm5, %%mm0                \n\t" // (ABS(block[i])*qmat[0] + bias[0]*qmat[0])>>16
 
172
            "por %%mm0, %%mm4                   \n\t"
 
173
            "pxor %%mm1, %%mm0                  \n\t"
 
174
            "psubw %%mm1, %%mm0                 \n\t" // out=((ABS(block[i])*qmat[0] - bias[0]*qmat[0])>>16)*sign(block[i])
 
175
            "movq %%mm0, (%5, %%"REG_a")        \n\t"
 
176
            "pcmpeqw %%mm7, %%mm0               \n\t" // out==0 ? 0xFF : 0x00
 
177
            "movq (%4, %%"REG_a"), %%mm1        \n\t"
 
178
            "movq %%mm7, (%1, %%"REG_a")        \n\t" // 0
 
179
            "pandn %%mm1, %%mm0                 \n\t"
 
180
            PMAXW(%%mm0, %%mm3)
 
181
            "add $8, %%"REG_a"                  \n\t"
 
182
            " js 1b                             \n\t"
 
183
            PMAX(%%mm3, %%mm0)
 
184
            "movd %%mm3, %%"REG_a"              \n\t"
 
185
            "movzb %%al, %%"REG_a"              \n\t" // last_non_zero_p1
 
186
            : "+a" (last_non_zero_p1)
178
187
            : "r" (block+64), "r" (qmat+64), "r" (bias+64),
179
188
              "r" (inv_zigzag_direct16+64), "r" (temp_block+64)
180
189
        );
181
190
        // note the asm is split cuz gcc doesnt like that many operands ...
182
191
        asm volatile(
183
 
            "movd %1, %%mm1                     \n\t" // max_qcoeff
184
 
            SPREADW(%%mm1)
185
 
            "psubusw %%mm1, %%mm4               \n\t" 
186
 
            "packuswb %%mm4, %%mm4              \n\t"
187
 
            "movd %%mm4, %0                     \n\t" // *overflow
 
192
            "movd %1, %%mm1                     \n\t" // max_qcoeff
 
193
            SPREADW(%%mm1)
 
194
            "psubusw %%mm1, %%mm4               \n\t"
 
195
            "packuswb %%mm4, %%mm4              \n\t"
 
196
            "movd %%mm4, %0                     \n\t" // *overflow
188
197
        : "=g" (*overflow)
189
198
        : "g" (s->max_qcoeff)
190
199
        );
195
204
 
196
205
    if(s->dsp.idct_permutation_type == FF_SIMPLE_IDCT_PERM){
197
206
        if(last_non_zero_p1 <= 1) goto end;
198
 
        block[0x08] = temp_block[0x01]; block[0x10] = temp_block[0x08]; 
199
 
        block[0x20] = temp_block[0x10]; 
 
207
        block[0x08] = temp_block[0x01]; block[0x10] = temp_block[0x08];
 
208
        block[0x20] = temp_block[0x10];
200
209
        if(last_non_zero_p1 <= 4) goto end;
201
 
        block[0x18] = temp_block[0x09]; block[0x04] = temp_block[0x02]; 
202
 
        block[0x09] = temp_block[0x03]; 
 
210
        block[0x18] = temp_block[0x09]; block[0x04] = temp_block[0x02];
 
211
        block[0x09] = temp_block[0x03];
203
212
        if(last_non_zero_p1 <= 7) goto end;
204
 
        block[0x14] = temp_block[0x0A]; block[0x28] = temp_block[0x11]; 
205
 
        block[0x12] = temp_block[0x18]; block[0x02] = temp_block[0x20]; 
 
213
        block[0x14] = temp_block[0x0A]; block[0x28] = temp_block[0x11];
 
214
        block[0x12] = temp_block[0x18]; block[0x02] = temp_block[0x20];
206
215
        if(last_non_zero_p1 <= 11) goto end;
207
 
        block[0x1A] = temp_block[0x19]; block[0x24] = temp_block[0x12]; 
208
 
        block[0x19] = temp_block[0x0B]; block[0x01] = temp_block[0x04]; 
209
 
        block[0x0C] = temp_block[0x05]; 
 
216
        block[0x1A] = temp_block[0x19]; block[0x24] = temp_block[0x12];
 
217
        block[0x19] = temp_block[0x0B]; block[0x01] = temp_block[0x04];
 
218
        block[0x0C] = temp_block[0x05];
210
219
        if(last_non_zero_p1 <= 16) goto end;
211
 
        block[0x11] = temp_block[0x0C]; block[0x29] = temp_block[0x13]; 
212
 
        block[0x16] = temp_block[0x1A]; block[0x0A] = temp_block[0x21]; 
213
 
        block[0x30] = temp_block[0x28]; block[0x22] = temp_block[0x30]; 
214
 
        block[0x38] = temp_block[0x29]; block[0x06] = temp_block[0x22]; 
 
220
        block[0x11] = temp_block[0x0C]; block[0x29] = temp_block[0x13];
 
221
        block[0x16] = temp_block[0x1A]; block[0x0A] = temp_block[0x21];
 
222
        block[0x30] = temp_block[0x28]; block[0x22] = temp_block[0x30];
 
223
        block[0x38] = temp_block[0x29]; block[0x06] = temp_block[0x22];
215
224
        if(last_non_zero_p1 <= 24) goto end;
216
 
        block[0x1B] = temp_block[0x1B]; block[0x21] = temp_block[0x14]; 
217
 
        block[0x1C] = temp_block[0x0D]; block[0x05] = temp_block[0x06]; 
218
 
        block[0x0D] = temp_block[0x07]; block[0x15] = temp_block[0x0E]; 
219
 
        block[0x2C] = temp_block[0x15]; block[0x13] = temp_block[0x1C]; 
 
225
        block[0x1B] = temp_block[0x1B]; block[0x21] = temp_block[0x14];
 
226
        block[0x1C] = temp_block[0x0D]; block[0x05] = temp_block[0x06];
 
227
        block[0x0D] = temp_block[0x07]; block[0x15] = temp_block[0x0E];
 
228
        block[0x2C] = temp_block[0x15]; block[0x13] = temp_block[0x1C];
220
229
        if(last_non_zero_p1 <= 32) goto end;
221
 
        block[0x0B] = temp_block[0x23]; block[0x34] = temp_block[0x2A]; 
222
 
        block[0x2A] = temp_block[0x31]; block[0x32] = temp_block[0x38]; 
223
 
        block[0x3A] = temp_block[0x39]; block[0x26] = temp_block[0x32]; 
224
 
        block[0x39] = temp_block[0x2B]; block[0x03] = temp_block[0x24]; 
 
230
        block[0x0B] = temp_block[0x23]; block[0x34] = temp_block[0x2A];
 
231
        block[0x2A] = temp_block[0x31]; block[0x32] = temp_block[0x38];
 
232
        block[0x3A] = temp_block[0x39]; block[0x26] = temp_block[0x32];
 
233
        block[0x39] = temp_block[0x2B]; block[0x03] = temp_block[0x24];
225
234
        if(last_non_zero_p1 <= 40) goto end;
226
 
        block[0x1E] = temp_block[0x1D]; block[0x25] = temp_block[0x16]; 
227
 
        block[0x1D] = temp_block[0x0F]; block[0x2D] = temp_block[0x17]; 
228
 
        block[0x17] = temp_block[0x1E]; block[0x0E] = temp_block[0x25]; 
229
 
        block[0x31] = temp_block[0x2C]; block[0x2B] = temp_block[0x33]; 
 
235
        block[0x1E] = temp_block[0x1D]; block[0x25] = temp_block[0x16];
 
236
        block[0x1D] = temp_block[0x0F]; block[0x2D] = temp_block[0x17];
 
237
        block[0x17] = temp_block[0x1E]; block[0x0E] = temp_block[0x25];
 
238
        block[0x31] = temp_block[0x2C]; block[0x2B] = temp_block[0x33];
230
239
        if(last_non_zero_p1 <= 48) goto end;
231
 
        block[0x36] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; 
232
 
        block[0x23] = temp_block[0x34]; block[0x3C] = temp_block[0x2D]; 
233
 
        block[0x07] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; 
234
 
        block[0x0F] = temp_block[0x27]; block[0x35] = temp_block[0x2E]; 
 
240
        block[0x36] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B];
 
241
        block[0x23] = temp_block[0x34]; block[0x3C] = temp_block[0x2D];
 
242
        block[0x07] = temp_block[0x26]; block[0x1F] = temp_block[0x1F];
 
243
        block[0x0F] = temp_block[0x27]; block[0x35] = temp_block[0x2E];
235
244
        if(last_non_zero_p1 <= 56) goto end;
236
 
        block[0x2E] = temp_block[0x35]; block[0x33] = temp_block[0x3C]; 
237
 
        block[0x3E] = temp_block[0x3D]; block[0x27] = temp_block[0x36]; 
238
 
        block[0x3D] = temp_block[0x2F]; block[0x2F] = temp_block[0x37]; 
 
245
        block[0x2E] = temp_block[0x35]; block[0x33] = temp_block[0x3C];
 
246
        block[0x3E] = temp_block[0x3D]; block[0x27] = temp_block[0x36];
 
247
        block[0x3D] = temp_block[0x2F]; block[0x2F] = temp_block[0x37];
239
248
        block[0x37] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F];
240
249
    }else if(s->dsp.idct_permutation_type == FF_LIBMPEG2_IDCT_PERM){
241
250
        if(last_non_zero_p1 <= 1) goto end;
242
 
        block[0x04] = temp_block[0x01]; 
243
 
        block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; 
 
251
        block[0x04] = temp_block[0x01];
 
252
        block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10];
244
253
        if(last_non_zero_p1 <= 4) goto end;
245
 
        block[0x0C] = temp_block[0x09]; block[0x01] = temp_block[0x02]; 
246
 
        block[0x05] = temp_block[0x03]; 
 
254
        block[0x0C] = temp_block[0x09]; block[0x01] = temp_block[0x02];
 
255
        block[0x05] = temp_block[0x03];
247
256
        if(last_non_zero_p1 <= 7) goto end;
248
 
        block[0x09] = temp_block[0x0A]; block[0x14] = temp_block[0x11]; 
249
 
        block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; 
 
257
        block[0x09] = temp_block[0x0A]; block[0x14] = temp_block[0x11];
 
258
        block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20];
250
259
        if(last_non_zero_p1 <= 11) goto end;
251
 
        block[0x1C] = temp_block[0x19]; 
252
 
        block[0x11] = temp_block[0x12]; block[0x0D] = temp_block[0x0B]; 
253
 
        block[0x02] = temp_block[0x04]; block[0x06] = temp_block[0x05]; 
 
260
        block[0x1C] = temp_block[0x19];
 
261
        block[0x11] = temp_block[0x12]; block[0x0D] = temp_block[0x0B];
 
262
        block[0x02] = temp_block[0x04]; block[0x06] = temp_block[0x05];
254
263
        if(last_non_zero_p1 <= 16) goto end;
255
 
        block[0x0A] = temp_block[0x0C]; block[0x15] = temp_block[0x13]; 
256
 
        block[0x19] = temp_block[0x1A]; block[0x24] = temp_block[0x21]; 
257
 
        block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; 
258
 
        block[0x2C] = temp_block[0x29]; block[0x21] = temp_block[0x22]; 
 
264
        block[0x0A] = temp_block[0x0C]; block[0x15] = temp_block[0x13];
 
265
        block[0x19] = temp_block[0x1A]; block[0x24] = temp_block[0x21];
 
266
        block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30];
 
267
        block[0x2C] = temp_block[0x29]; block[0x21] = temp_block[0x22];
259
268
        if(last_non_zero_p1 <= 24) goto end;
260
 
        block[0x1D] = temp_block[0x1B]; block[0x12] = temp_block[0x14]; 
261
 
        block[0x0E] = temp_block[0x0D]; block[0x03] = temp_block[0x06]; 
262
 
        block[0x07] = temp_block[0x07]; block[0x0B] = temp_block[0x0E]; 
263
 
        block[0x16] = temp_block[0x15]; block[0x1A] = temp_block[0x1C]; 
 
269
        block[0x1D] = temp_block[0x1B]; block[0x12] = temp_block[0x14];
 
270
        block[0x0E] = temp_block[0x0D]; block[0x03] = temp_block[0x06];
 
271
        block[0x07] = temp_block[0x07]; block[0x0B] = temp_block[0x0E];
 
272
        block[0x16] = temp_block[0x15]; block[0x1A] = temp_block[0x1C];
264
273
        if(last_non_zero_p1 <= 32) goto end;
265
 
        block[0x25] = temp_block[0x23]; block[0x29] = temp_block[0x2A]; 
266
 
        block[0x34] = temp_block[0x31]; block[0x38] = temp_block[0x38]; 
267
 
        block[0x3C] = temp_block[0x39]; block[0x31] = temp_block[0x32]; 
268
 
        block[0x2D] = temp_block[0x2B]; block[0x22] = temp_block[0x24]; 
 
274
        block[0x25] = temp_block[0x23]; block[0x29] = temp_block[0x2A];
 
275
        block[0x34] = temp_block[0x31]; block[0x38] = temp_block[0x38];
 
276
        block[0x3C] = temp_block[0x39]; block[0x31] = temp_block[0x32];
 
277
        block[0x2D] = temp_block[0x2B]; block[0x22] = temp_block[0x24];
269
278
        if(last_non_zero_p1 <= 40) goto end;
270
 
        block[0x1E] = temp_block[0x1D]; block[0x13] = temp_block[0x16]; 
271
 
        block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; 
272
 
        block[0x1B] = temp_block[0x1E]; block[0x26] = temp_block[0x25]; 
273
 
        block[0x2A] = temp_block[0x2C]; block[0x35] = temp_block[0x33]; 
 
279
        block[0x1E] = temp_block[0x1D]; block[0x13] = temp_block[0x16];
 
280
        block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17];
 
281
        block[0x1B] = temp_block[0x1E]; block[0x26] = temp_block[0x25];
 
282
        block[0x2A] = temp_block[0x2C]; block[0x35] = temp_block[0x33];
274
283
        if(last_non_zero_p1 <= 48) goto end;
275
 
        block[0x39] = temp_block[0x3A]; block[0x3D] = temp_block[0x3B]; 
276
 
        block[0x32] = temp_block[0x34]; block[0x2E] = temp_block[0x2D]; 
277
 
            block[0x23] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; 
278
 
        block[0x27] = temp_block[0x27]; block[0x2B] = temp_block[0x2E]; 
 
284
        block[0x39] = temp_block[0x3A]; block[0x3D] = temp_block[0x3B];
 
285
        block[0x32] = temp_block[0x34]; block[0x2E] = temp_block[0x2D];
 
286
            block[0x23] = temp_block[0x26]; block[0x1F] = temp_block[0x1F];
 
287
        block[0x27] = temp_block[0x27]; block[0x2B] = temp_block[0x2E];
279
288
        if(last_non_zero_p1 <= 56) goto end;
280
 
        block[0x36] = temp_block[0x35]; block[0x3A] = temp_block[0x3C]; 
281
 
        block[0x3E] = temp_block[0x3D]; block[0x33] = temp_block[0x36]; 
282
 
        block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; 
 
289
        block[0x36] = temp_block[0x35]; block[0x3A] = temp_block[0x3C];
 
290
        block[0x3E] = temp_block[0x3D]; block[0x33] = temp_block[0x36];
 
291
        block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37];
283
292
        block[0x3B] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F];
284
293
    }else{
285
294
        if(last_non_zero_p1 <= 1) goto end;
286
 
        block[0x01] = temp_block[0x01]; 
287
 
        block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10]; 
 
295
        block[0x01] = temp_block[0x01];
 
296
        block[0x08] = temp_block[0x08]; block[0x10] = temp_block[0x10];
288
297
        if(last_non_zero_p1 <= 4) goto end;
289
 
        block[0x09] = temp_block[0x09]; block[0x02] = temp_block[0x02]; 
290
 
        block[0x03] = temp_block[0x03]; 
 
298
        block[0x09] = temp_block[0x09]; block[0x02] = temp_block[0x02];
 
299
        block[0x03] = temp_block[0x03];
291
300
        if(last_non_zero_p1 <= 7) goto end;
292
 
        block[0x0A] = temp_block[0x0A]; block[0x11] = temp_block[0x11]; 
293
 
        block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20]; 
 
301
        block[0x0A] = temp_block[0x0A]; block[0x11] = temp_block[0x11];
 
302
        block[0x18] = temp_block[0x18]; block[0x20] = temp_block[0x20];
294
303
        if(last_non_zero_p1 <= 11) goto end;
295
 
        block[0x19] = temp_block[0x19]; 
296
 
        block[0x12] = temp_block[0x12]; block[0x0B] = temp_block[0x0B]; 
297
 
        block[0x04] = temp_block[0x04]; block[0x05] = temp_block[0x05]; 
 
304
        block[0x19] = temp_block[0x19];
 
305
        block[0x12] = temp_block[0x12]; block[0x0B] = temp_block[0x0B];
 
306
        block[0x04] = temp_block[0x04]; block[0x05] = temp_block[0x05];
298
307
        if(last_non_zero_p1 <= 16) goto end;
299
 
        block[0x0C] = temp_block[0x0C]; block[0x13] = temp_block[0x13]; 
300
 
        block[0x1A] = temp_block[0x1A]; block[0x21] = temp_block[0x21]; 
301
 
        block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30]; 
302
 
        block[0x29] = temp_block[0x29]; block[0x22] = temp_block[0x22]; 
 
308
        block[0x0C] = temp_block[0x0C]; block[0x13] = temp_block[0x13];
 
309
        block[0x1A] = temp_block[0x1A]; block[0x21] = temp_block[0x21];
 
310
        block[0x28] = temp_block[0x28]; block[0x30] = temp_block[0x30];
 
311
        block[0x29] = temp_block[0x29]; block[0x22] = temp_block[0x22];
303
312
        if(last_non_zero_p1 <= 24) goto end;
304
 
        block[0x1B] = temp_block[0x1B]; block[0x14] = temp_block[0x14]; 
305
 
        block[0x0D] = temp_block[0x0D]; block[0x06] = temp_block[0x06]; 
306
 
        block[0x07] = temp_block[0x07]; block[0x0E] = temp_block[0x0E]; 
307
 
        block[0x15] = temp_block[0x15]; block[0x1C] = temp_block[0x1C]; 
 
313
        block[0x1B] = temp_block[0x1B]; block[0x14] = temp_block[0x14];
 
314
        block[0x0D] = temp_block[0x0D]; block[0x06] = temp_block[0x06];
 
315
        block[0x07] = temp_block[0x07]; block[0x0E] = temp_block[0x0E];
 
316
        block[0x15] = temp_block[0x15]; block[0x1C] = temp_block[0x1C];
308
317
        if(last_non_zero_p1 <= 32) goto end;
309
 
        block[0x23] = temp_block[0x23]; block[0x2A] = temp_block[0x2A]; 
310
 
        block[0x31] = temp_block[0x31]; block[0x38] = temp_block[0x38]; 
311
 
        block[0x39] = temp_block[0x39]; block[0x32] = temp_block[0x32]; 
312
 
        block[0x2B] = temp_block[0x2B]; block[0x24] = temp_block[0x24]; 
 
318
        block[0x23] = temp_block[0x23]; block[0x2A] = temp_block[0x2A];
 
319
        block[0x31] = temp_block[0x31]; block[0x38] = temp_block[0x38];
 
320
        block[0x39] = temp_block[0x39]; block[0x32] = temp_block[0x32];
 
321
        block[0x2B] = temp_block[0x2B]; block[0x24] = temp_block[0x24];
313
322
        if(last_non_zero_p1 <= 40) goto end;
314
 
        block[0x1D] = temp_block[0x1D]; block[0x16] = temp_block[0x16]; 
315
 
        block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17]; 
316
 
        block[0x1E] = temp_block[0x1E]; block[0x25] = temp_block[0x25]; 
317
 
        block[0x2C] = temp_block[0x2C]; block[0x33] = temp_block[0x33]; 
 
323
        block[0x1D] = temp_block[0x1D]; block[0x16] = temp_block[0x16];
 
324
        block[0x0F] = temp_block[0x0F]; block[0x17] = temp_block[0x17];
 
325
        block[0x1E] = temp_block[0x1E]; block[0x25] = temp_block[0x25];
 
326
        block[0x2C] = temp_block[0x2C]; block[0x33] = temp_block[0x33];
318
327
        if(last_non_zero_p1 <= 48) goto end;
319
 
        block[0x3A] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B]; 
320
 
        block[0x34] = temp_block[0x34]; block[0x2D] = temp_block[0x2D]; 
321
 
        block[0x26] = temp_block[0x26]; block[0x1F] = temp_block[0x1F]; 
322
 
        block[0x27] = temp_block[0x27]; block[0x2E] = temp_block[0x2E]; 
 
328
        block[0x3A] = temp_block[0x3A]; block[0x3B] = temp_block[0x3B];
 
329
        block[0x34] = temp_block[0x34]; block[0x2D] = temp_block[0x2D];
 
330
        block[0x26] = temp_block[0x26]; block[0x1F] = temp_block[0x1F];
 
331
        block[0x27] = temp_block[0x27]; block[0x2E] = temp_block[0x2E];
323
332
        if(last_non_zero_p1 <= 56) goto end;
324
 
        block[0x35] = temp_block[0x35]; block[0x3C] = temp_block[0x3C]; 
325
 
        block[0x3D] = temp_block[0x3D]; block[0x36] = temp_block[0x36]; 
326
 
        block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37]; 
 
333
        block[0x35] = temp_block[0x35]; block[0x3C] = temp_block[0x3C];
 
334
        block[0x3D] = temp_block[0x3D]; block[0x36] = temp_block[0x36];
 
335
        block[0x2F] = temp_block[0x2F]; block[0x37] = temp_block[0x37];
327
336
        block[0x3E] = temp_block[0x3E]; block[0x3F] = temp_block[0x3F];
328
337
    }
329
338
    end: