← Back to branch summary

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

« back to all changes in this revision

Viewing changes to ffmpeg/libavcodec/sh4/idct_sh4.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/sh4/idct_sh4.c
3
3
 *
4
4
 * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
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
 
21
23
#include "../dsputil.h"
22
 
#define c1      1.38703984532214752434  /* sqrt(2)*cos(1*pi/16) */
23
 
#define c2      1.30656296487637657577  /* sqrt(2)*cos(2*pi/16) */
24
 
#define c3      1.17587560241935884520  /* sqrt(2)*cos(3*pi/16) */
25
 
#define c4      1.00000000000000000000  /* sqrt(2)*cos(4*pi/16) */
26
 
#define c5      0.78569495838710234903  /* sqrt(2)*cos(5*pi/16) */
27
 
#define c6      0.54119610014619712324  /* sqrt(2)*cos(6*pi/16) */
28
 
#define c7      0.27589937928294311353  /* sqrt(2)*cos(7*pi/16) */
29
 
 
30
 
const   static float even_table[] __attribute__ ((aligned(8))) = {
31
 
         c4, c4, c4, c4,
32
 
         c2, c6,-c6,-c2,
33
 
         c4,-c4,-c4, c4,
34
 
         c6,-c2, c2,-c6
35
 
};
36
 
 
37
 
const   static float odd_table[] __attribute__ ((aligned(8))) = {
38
 
         c1, c3, c5, c7,
39
 
         c3,-c7,-c1,-c5,
40
 
         c5,-c1, c7, c3,
41
 
         c7,-c5, c3,-c1
42
 
};
43
 
 
44
 
#undef  c1
45
 
#undef  c2
46
 
#undef  c3
47
 
#undef  c4
48
 
#undef  c5
49
 
#undef  c6
50
 
#undef  c7
 
24
#define c1      1.38703984532214752434  /* sqrt(2)*cos(1*pi/16) */
 
25
#define c2      1.30656296487637657577  /* sqrt(2)*cos(2*pi/16) */
 
26
#define c3      1.17587560241935884520  /* sqrt(2)*cos(3*pi/16) */
 
27
#define c4      1.00000000000000000000  /* sqrt(2)*cos(4*pi/16) */
 
28
#define c5      0.78569495838710234903  /* sqrt(2)*cos(5*pi/16) */
 
29
#define c6      0.54119610014619712324  /* sqrt(2)*cos(6*pi/16) */
 
30
#define c7      0.27589937928294311353  /* sqrt(2)*cos(7*pi/16) */
 
31
 
 
32
static const float even_table[] __attribute__ ((aligned(8))) = {
 
33
        c4, c4, c4, c4,
 
34
        c2, c6,-c6,-c2,
 
35
        c4,-c4,-c4, c4,
 
36
        c6,-c2, c2,-c6
 
37
};
 
38
 
 
39
static const float odd_table[] __attribute__ ((aligned(8))) = {
 
40
        c1, c3, c5, c7,
 
41
        c3,-c7,-c1,-c5,
 
42
        c5,-c1, c7, c3,
 
43
        c7,-c5, c3,-c1
 
44
};
 
45
 
 
46
#undef  c1
 
47
#undef  c2
 
48
#undef  c3
 
49
#undef  c4
 
50
#undef  c5
 
51
#undef  c6
 
52
#undef  c7
51
53
 
52
54
#if defined(__SH4_SINGLE__) || defined(__SH4_SINGLE_ONLY__)
53
55
 
54
 
#define load_matrix(table) \
55
 
        __asm__ volatile( \
56
 
        "       fschg\n" \
57
 
        "       fmov    @%0+,xd0\n" \
58
 
        "       fmov    @%0+,xd2\n" \
59
 
        "       fmov    @%0+,xd4\n" \
60
 
        "       fmov    @%0+,xd6\n" \
61
 
        "       fmov    @%0+,xd8\n" \
62
 
        "       fmov    @%0+,xd10\n" \
63
 
        "       fmov    @%0+,xd12\n" \
64
 
        "       fmov    @%0+,xd14\n" \
65
 
        "       fschg\n" \
66
 
        :\
67
 
        : "r"(table)\
68
 
        : "0" \
69
 
        )
70
 
 
71
 
#define ftrv() \
72
 
                __asm__ volatile("ftrv xmtrx,fv0" \
73
 
                : "=f"(fr0),"=f"(fr1),"=f"(fr2),"=f"(fr3) \
74
 
                :  "0"(fr0), "1"(fr1), "2"(fr2), "3"(fr3) );
75
 
 
76
 
#define DEFREG  \
77
 
        register float fr0 __asm__("fr0"); \
78
 
        register float fr1 __asm__("fr1"); \
79
 
        register float fr2 __asm__("fr2"); \
80
 
        register float fr3 __asm__("fr3")
 
56
#define         load_matrix(table) \
 
57
        __asm__ volatile( \
 
58
        "       fschg\n" \
 
59
        "       fmov   @%0+,xd0\n" \
 
60
        "       fmov   @%0+,xd2\n" \
 
61
        "       fmov   @%0+,xd4\n" \
 
62
        "       fmov   @%0+,xd6\n" \
 
63
        "       fmov   @%0+,xd8\n" \
 
64
        "       fmov   @%0+,xd10\n" \
 
65
        "       fmov   @%0+,xd12\n" \
 
66
        "       fmov   @%0+,xd14\n" \
 
67
        "       fschg\n" \
 
68
        :\
 
69
        : "r"(table)\
 
70
        : "0" \
 
71
        )
 
72
 
 
73
#define         ftrv() \
 
74
                __asm__ volatile("ftrv xmtrx,fv0" \
 
75
                : "=f"(fr0),"=f"(fr1),"=f"(fr2),"=f"(fr3) \
 
76
                :  "0"(fr0), "1"(fr1), "2"(fr2), "3"(fr3) );
 
77
 
 
78
#define         DEFREG        \
 
79
        register float fr0 __asm__("fr0"); \
 
80
        register float fr1 __asm__("fr1"); \
 
81
        register float fr2 __asm__("fr2"); \
 
82
        register float fr3 __asm__("fr3")
81
83
 
82
84
#else
83
85
 
85
87
 
86
88
static void ftrv_(const float xf[],float fv[])
87
89
{
88
 
        float f0,f1,f2,f3;
89
 
        f0 = fv[0];
90
 
        f1 = fv[1];
91
 
        f2 = fv[2];
92
 
        f3 = fv[3];
93
 
        fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3;
94
 
        fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3;
95
 
        fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3;
96
 
        fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3;
 
90
        float f0,f1,f2,f3;
 
91
        f0 = fv[0];
 
92
        f1 = fv[1];
 
93
        f2 = fv[2];
 
94
        f3 = fv[3];
 
95
        fv[0] = xf[0]*f0 + xf[4]*f1 + xf[ 8]*f2 + xf[12]*f3;
 
96
        fv[1] = xf[1]*f0 + xf[5]*f1 + xf[ 9]*f2 + xf[13]*f3;
 
97
        fv[2] = xf[2]*f0 + xf[6]*f1 + xf[10]*f2 + xf[14]*f3;
 
98
        fv[3] = xf[3]*f0 + xf[7]*f1 + xf[11]*f2 + xf[15]*f3;
97
99
}
98
100
 
99
101
static void load_matrix_(float xf[],const float table[])
100
102
{
101
 
        int i;
102
 
        for(i=0;i<16;i++) xf[i]=table[i];
 
103
        int i;
 
104
        for(i=0;i<16;i++) xf[i]=table[i];
103
105
}
104
106
 
105
 
#define ftrv()  ftrv_(xf,fv)
106
 
#define load_matrix(table)      load_matrix_(xf,table)
107
 
 
108
 
#define DEFREG \
109
 
        float fv[4],xf[16]
110
 
 
111
 
#define fr0     fv[0]
112
 
#define fr1     fv[1]
113
 
#define fr2     fv[2]
114
 
#define fr3     fv[3]
 
107
#define         ftrv()                  ftrv_(xf,fv)
 
108
#define         load_matrix(table)      load_matrix_(xf,table)
 
109
 
 
110
#define         DEFREG \
 
111
        float fv[4],xf[16]
 
112
 
 
113
#define         fr0     fv[0]
 
114
#define         fr1     fv[1]
 
115
#define         fr2     fv[2]
 
116
#define         fr3     fv[3]
115
117
 
116
118
#endif
117
119
 
118
120
#if 1
119
 
#define DESCALE(x,n)    (x)*(1.0f/(1<<(n)))
 
121
#define         DESCALE(x,n)    (x)*(1.0f/(1<<(n)))
120
122
#else
121
 
#define DESCALE(x,n)    (((int)(x)+(1<<(n-1)))>>(n))
 
123
#define         DESCALE(x,n)    (((int)(x)+(1<<(n-1)))>>(n))
122
124
#endif
123
125
 
124
126
/* this code work worse on gcc cvs. 3.2.3 work fine */
125
127
 
126
128
 
127
129
#if 1
128
 
//optimized 
 
130
//optimized
129
131
 
130
132
void idct_sh4(DCTELEM *block)
131
133
{
132
 
        DEFREG;
 
134
        DEFREG;
133
135
 
134
 
        int i;
135
 
        float   tblock[8*8],*fblock;
136
 
        int ofs1,ofs2,ofs3;
 
136
        int i;
 
137
        float        tblock[8*8],*fblock;
 
138
        int ofs1,ofs2,ofs3;
137
139
 
138
140
#if defined(__SH4__)
139
 
#error  "FIXME!! change to single float"
 
141
#error  "FIXME!! change to single float"
140
142
#endif
141
143
 
142
 
        /* row */
143
 
 
144
 
        /* even part */
145
 
        load_matrix(even_table);
146
 
 
147
 
        fblock = tblock+4;
148
 
        i = 8;
149
 
        do {
150
 
                fr0 = block[0];
151
 
                fr1 = block[2];
152
 
                fr2 = block[4];
153
 
                fr3 = block[6];
154
 
                block+=8;
155
 
                ftrv();
156
 
                *--fblock = fr3;
157
 
                *--fblock = fr2;
158
 
                *--fblock = fr1;
159
 
                *--fblock = fr0;
160
 
                fblock+=8+4;
161
 
        } while(--i);
162
 
        block-=8*8;
163
 
        fblock-=8*8+4;
164
 
 
165
 
        load_matrix(odd_table);
166
 
 
167
 
        i = 8;
168
 
 
169
 
//      ofs1 = sizeof(float)*1;
170
 
//      ofs2 = sizeof(float)*2;
171
 
//      ofs3 = sizeof(float)*3;
172
 
 
173
 
        do {
174
 
                float t0,t1,t2,t3;
175
 
                fr0 = block[1];
176
 
                fr1 = block[3];
177
 
                fr2 = block[5];
178
 
                fr3 = block[7];
179
 
                block+=8;
180
 
                ftrv();
181
 
                t0 = *fblock++;
182
 
                t1 = *fblock++;
183
 
                t2 = *fblock++;
184
 
                t3 = *fblock++;
185
 
                fblock+=4;
186
 
                *--fblock = t0 - fr0;
187
 
                *--fblock = t1 - fr1;
188
 
                *--fblock = t2 - fr2;
189
 
                *--fblock = t3 - fr3;
190
 
                *--fblock = t3 + fr3;
191
 
                *--fblock = t2 + fr2;
192
 
                *--fblock = t1 + fr1;
193
 
                *--fblock = t0 + fr0;
194
 
                fblock+=8;
195
 
        } while(--i);
196
 
        block-=8*8;
197
 
        fblock-=8*8;
198
 
 
199
 
        /* col */
200
 
 
201
 
        /* even part */
202
 
        load_matrix(even_table);
203
 
 
204
 
        ofs1 = sizeof(float)*2*8;
205
 
        ofs2 = sizeof(float)*4*8;
206
 
        ofs3 = sizeof(float)*6*8;
207
 
 
208
 
        i = 8;
209
 
 
210
 
#define OA(fblock,ofs)  *(float*)((char*)fblock + ofs)
211
 
 
212
 
        do {
213
 
                fr0 = OA(fblock,   0);
214
 
                fr1 = OA(fblock,ofs1);
215
 
                fr2 = OA(fblock,ofs2);
216
 
                fr3 = OA(fblock,ofs3);
217
 
                ftrv();
218
 
                OA(fblock,0   ) = fr0;
219
 
                OA(fblock,ofs1) = fr1;
220
 
                OA(fblock,ofs2) = fr2;
221
 
                OA(fblock,ofs3) = fr3;
222
 
                fblock++;
223
 
        } while(--i);
224
 
        fblock-=8;
225
 
 
226
 
        load_matrix(odd_table);
227
 
 
228
 
        i=8;
229
 
        do {
230
 
                float t0,t1,t2,t3;
231
 
                t0 = OA(fblock,   0); /* [8*0] */
232
 
                t1 = OA(fblock,ofs1); /* [8*2] */
233
 
                t2 = OA(fblock,ofs2); /* [8*4] */
234
 
                t3 = OA(fblock,ofs3); /* [8*6] */
235
 
                fblock+=8;
236
 
                fr0 = OA(fblock,   0); /* [8*1] */
237
 
                fr1 = OA(fblock,ofs1); /* [8*3] */
238
 
                fr2 = OA(fblock,ofs2); /* [8*5] */
239
 
                fr3 = OA(fblock,ofs3); /* [8*7] */
240
 
                fblock+=-8+1;
241
 
                ftrv();
242
 
                block[8*0] = DESCALE(t0 + fr0,3);
243
 
                block[8*7] = DESCALE(t0 - fr0,3);
244
 
                block[8*1] = DESCALE(t1 + fr1,3);
245
 
                block[8*6] = DESCALE(t1 - fr1,3);
246
 
                block[8*2] = DESCALE(t2 + fr2,3);
247
 
                block[8*5] = DESCALE(t2 - fr2,3);
248
 
                block[8*3] = DESCALE(t3 + fr3,3);
249
 
                block[8*4] = DESCALE(t3 - fr3,3);
250
 
                block++;
251
 
        } while(--i);
 
144
        /* row */
 
145
 
 
146
        /* even part */
 
147
        load_matrix(even_table);
 
148
 
 
149
        fblock = tblock+4;
 
150
        i = 8;
 
151
        do {
 
152
                fr0 = block[0];
 
153
                fr1 = block[2];
 
154
                fr2 = block[4];
 
155
                fr3 = block[6];
 
156
                block+=8;
 
157
                ftrv();
 
158
                *--fblock = fr3;
 
159
                *--fblock = fr2;
 
160
                *--fblock = fr1;
 
161
                *--fblock = fr0;
 
162
                fblock+=8+4;
 
163
        } while(--i);
 
164
        block-=8*8;
 
165
        fblock-=8*8+4;
 
166
 
 
167
        load_matrix(odd_table);
 
168
 
 
169
        i = 8;
 
170
 
 
171
//        ofs1 = sizeof(float)*1;
 
172
//        ofs2 = sizeof(float)*2;
 
173
//        ofs3 = sizeof(float)*3;
 
174
 
 
175
        do {
 
176
                float t0,t1,t2,t3;
 
177
                fr0 = block[1];
 
178
                fr1 = block[3];
 
179
                fr2 = block[5];
 
180
                fr3 = block[7];
 
181
                block+=8;
 
182
                ftrv();
 
183
                t0 = *fblock++;
 
184
                t1 = *fblock++;
 
185
                t2 = *fblock++;
 
186
                t3 = *fblock++;
 
187
                fblock+=4;
 
188
                *--fblock = t0 - fr0;
 
189
                *--fblock = t1 - fr1;
 
190
                *--fblock = t2 - fr2;
 
191
                *--fblock = t3 - fr3;
 
192
                *--fblock = t3 + fr3;
 
193
                *--fblock = t2 + fr2;
 
194
                *--fblock = t1 + fr1;
 
195
                *--fblock = t0 + fr0;
 
196
                fblock+=8;
 
197
        } while(--i);
 
198
        block-=8*8;
 
199
        fblock-=8*8;
 
200
 
 
201
        /* col */
 
202
 
 
203
        /* even part */
 
204
        load_matrix(even_table);
 
205
 
 
206
        ofs1 = sizeof(float)*2*8;
 
207
        ofs2 = sizeof(float)*4*8;
 
208
        ofs3 = sizeof(float)*6*8;
 
209
 
 
210
        i = 8;
 
211
 
 
212
#define        OA(fblock,ofs)   *(float*)((char*)fblock + ofs)
 
213
 
 
214
        do {
 
215
                fr0 = OA(fblock,   0);
 
216
                fr1 = OA(fblock,ofs1);
 
217
                fr2 = OA(fblock,ofs2);
 
218
                fr3 = OA(fblock,ofs3);
 
219
                ftrv();
 
220
                OA(fblock,0   ) = fr0;
 
221
                OA(fblock,ofs1) = fr1;
 
222
                OA(fblock,ofs2) = fr2;
 
223
                OA(fblock,ofs3) = fr3;
 
224
                fblock++;
 
225
        } while(--i);
 
226
        fblock-=8;
 
227
 
 
228
        load_matrix(odd_table);
 
229
 
 
230
        i=8;
 
231
        do {
 
232
                float t0,t1,t2,t3;
 
233
                t0 = OA(fblock,   0); /* [8*0] */
 
234
                t1 = OA(fblock,ofs1); /* [8*2] */
 
235
                t2 = OA(fblock,ofs2); /* [8*4] */
 
236
                t3 = OA(fblock,ofs3); /* [8*6] */
 
237
                fblock+=8;
 
238
                fr0 = OA(fblock,   0); /* [8*1] */
 
239
                fr1 = OA(fblock,ofs1); /* [8*3] */
 
240
                fr2 = OA(fblock,ofs2); /* [8*5] */
 
241
                fr3 = OA(fblock,ofs3); /* [8*7] */
 
242
                fblock+=-8+1;
 
243
                ftrv();
 
244
                block[8*0] = DESCALE(t0 + fr0,3);
 
245
                block[8*7] = DESCALE(t0 - fr0,3);
 
246
                block[8*1] = DESCALE(t1 + fr1,3);
 
247
                block[8*6] = DESCALE(t1 - fr1,3);
 
248
                block[8*2] = DESCALE(t2 + fr2,3);
 
249
                block[8*5] = DESCALE(t2 - fr2,3);
 
250
                block[8*3] = DESCALE(t3 + fr3,3);
 
251
                block[8*4] = DESCALE(t3 - fr3,3);
 
252
                block++;
 
253
        } while(--i);
252
254
 
253
255
#if defined(__SH4__)
254
 
#error  "FIXME!! change to double"
 
256
#error  "FIXME!! change to double"
255
257
#endif
256
258
}
257
259
#else
258
260
void idct_sh4(DCTELEM *block)
259
261
{
260
 
        DEFREG;
261
 
 
262
 
        int i;
263
 
        float   tblock[8*8],*fblock;
264
 
 
265
 
        /* row */
266
 
 
267
 
        /* even part */
268
 
        load_matrix(even_table);
269
 
 
270
 
        fblock = tblock;
271
 
        i = 8;
272
 
        do {
273
 
                fr0 = block[0];
274
 
                fr1 = block[2];
275
 
                fr2 = block[4];
276
 
                fr3 = block[6];
277
 
                block+=8;
278
 
                ftrv();
279
 
                fblock[0] = fr0;
280
 
                fblock[2] = fr1;
281
 
                fblock[4] = fr2;
282
 
                fblock[6] = fr3;
283
 
                fblock+=8;
284
 
        } while(--i);
285
 
        block-=8*8;
286
 
        fblock-=8*8;
287
 
 
288
 
        load_matrix(odd_table);
289
 
 
290
 
        i = 8;
291
 
 
292
 
        do {
293
 
                float t0,t1,t2,t3;
294
 
                fr0 = block[1];
295
 
                fr1 = block[3];
296
 
                fr2 = block[5];
297
 
                fr3 = block[7];
298
 
                block+=8;
299
 
                ftrv();
300
 
                t0 = fblock[0];
301
 
                t1 = fblock[2];
302
 
                t2 = fblock[4];
303
 
                t3 = fblock[6];
304
 
                fblock[0] = t0 + fr0;
305
 
                fblock[7] = t0 - fr0;
306
 
                fblock[1] = t1 + fr1;
307
 
                fblock[6] = t1 - fr1;
308
 
                fblock[2] = t2 + fr2;
309
 
                fblock[5] = t2 - fr2;
310
 
                fblock[3] = t3 + fr3;
311
 
                fblock[4] = t3 - fr3;
312
 
                fblock+=8;
313
 
        } while(--i);
314
 
        block-=8*8;
315
 
        fblock-=8*8;
316
 
 
317
 
        /* col */
318
 
 
319
 
        /* even part */
320
 
        load_matrix(even_table);
321
 
 
322
 
        i = 8;
323
 
 
324
 
        do {
325
 
                fr0 = fblock[8*0];
326
 
                fr1 = fblock[8*2];
327
 
                fr2 = fblock[8*4];
328
 
                fr3 = fblock[8*6];
329
 
                ftrv();
330
 
                fblock[8*0] = fr0;
331
 
                fblock[8*2] = fr1;
332
 
                fblock[8*4] = fr2;
333
 
                fblock[8*6] = fr3;
334
 
                fblock++;
335
 
        } while(--i);
336
 
        fblock-=8;
337
 
 
338
 
        load_matrix(odd_table);
339
 
 
340
 
        i=8;
341
 
        do {
342
 
                float t0,t1,t2,t3;
343
 
                fr0 = fblock[8*1];
344
 
                fr1 = fblock[8*3];
345
 
                fr2 = fblock[8*5];
346
 
                fr3 = fblock[8*7];
347
 
                ftrv();
348
 
                t0 = fblock[8*0];
349
 
                t1 = fblock[8*2];
350
 
                t2 = fblock[8*4];
351
 
                t3 = fblock[8*6];
352
 
                fblock++;
353
 
                block[8*0] = DESCALE(t0 + fr0,3);
354
 
                block[8*7] = DESCALE(t0 - fr0,3);
355
 
                block[8*1] = DESCALE(t1 + fr1,3);
356
 
                block[8*6] = DESCALE(t1 - fr1,3);
357
 
                block[8*2] = DESCALE(t2 + fr2,3);
358
 
                block[8*5] = DESCALE(t2 - fr2,3);
359
 
                block[8*3] = DESCALE(t3 + fr3,3);
360
 
                block[8*4] = DESCALE(t3 - fr3,3);
361
 
                block++;
362
 
        } while(--i);
 
262
        DEFREG;
 
263
 
 
264
        int i;
 
265
        float   tblock[8*8],*fblock;
 
266
 
 
267
        /* row */
 
268
 
 
269
        /* even part */
 
270
        load_matrix(even_table);
 
271
 
 
272
        fblock = tblock;
 
273
        i = 8;
 
274
        do {
 
275
                fr0 = block[0];
 
276
                fr1 = block[2];
 
277
                fr2 = block[4];
 
278
                fr3 = block[6];
 
279
                block+=8;
 
280
                ftrv();
 
281
                fblock[0] = fr0;
 
282
                fblock[2] = fr1;
 
283
                fblock[4] = fr2;
 
284
                fblock[6] = fr3;
 
285
                fblock+=8;
 
286
        } while(--i);
 
287
        block-=8*8;
 
288
        fblock-=8*8;
 
289
 
 
290
        load_matrix(odd_table);
 
291
 
 
292
        i = 8;
 
293
 
 
294
        do {
 
295
                float t0,t1,t2,t3;
 
296
                fr0 = block[1];
 
297
                fr1 = block[3];
 
298
                fr2 = block[5];
 
299
                fr3 = block[7];
 
300
                block+=8;
 
301
                ftrv();
 
302
                t0 = fblock[0];
 
303
                t1 = fblock[2];
 
304
                t2 = fblock[4];
 
305
                t3 = fblock[6];
 
306
                fblock[0] = t0 + fr0;
 
307
                fblock[7] = t0 - fr0;
 
308
                fblock[1] = t1 + fr1;
 
309
                fblock[6] = t1 - fr1;
 
310
                fblock[2] = t2 + fr2;
 
311
                fblock[5] = t2 - fr2;
 
312
                fblock[3] = t3 + fr3;
 
313
                fblock[4] = t3 - fr3;
 
314
                fblock+=8;
 
315
        } while(--i);
 
316
        block-=8*8;
 
317
        fblock-=8*8;
 
318
 
 
319
        /* col */
 
320
 
 
321
        /* even part */
 
322
        load_matrix(even_table);
 
323
 
 
324
        i = 8;
 
325
 
 
326
        do {
 
327
                fr0 = fblock[8*0];
 
328
                fr1 = fblock[8*2];
 
329
                fr2 = fblock[8*4];
 
330
                fr3 = fblock[8*6];
 
331
                ftrv();
 
332
                fblock[8*0] = fr0;
 
333
                fblock[8*2] = fr1;
 
334
                fblock[8*4] = fr2;
 
335
                fblock[8*6] = fr3;
 
336
                fblock++;
 
337
        } while(--i);
 
338
        fblock-=8;
 
339
 
 
340
        load_matrix(odd_table);
 
341
 
 
342
        i=8;
 
343
        do {
 
344
                float t0,t1,t2,t3;
 
345
                fr0 = fblock[8*1];
 
346
                fr1 = fblock[8*3];
 
347
                fr2 = fblock[8*5];
 
348
                fr3 = fblock[8*7];
 
349
                ftrv();
 
350
                t0 = fblock[8*0];
 
351
                t1 = fblock[8*2];
 
352
                t2 = fblock[8*4];
 
353
                t3 = fblock[8*6];
 
354
                fblock++;
 
355
                block[8*0] = DESCALE(t0 + fr0,3);
 
356
                block[8*7] = DESCALE(t0 - fr0,3);
 
357
                block[8*1] = DESCALE(t1 + fr1,3);
 
358
                block[8*6] = DESCALE(t1 - fr1,3);
 
359
                block[8*2] = DESCALE(t2 + fr2,3);
 
360
                block[8*5] = DESCALE(t2 - fr2,3);
 
361
                block[8*3] = DESCALE(t3 + fr3,3);
 
362
                block[8*4] = DESCALE(t3 - fr3,3);
 
363
                block++;
 
364
        } while(--i);
363
365
}
364
366
#endif