Statistics
| Branch: | Revision:

ffmpeg / libavcodec / dsputil.c @ ea937d01

History | View | Annotate | Download (78 KB)

1
/*
2
 * DSP utils
3
 * Copyright (c) 2000, 2001 Fabrice Bellard.
4
 *
5
 * This library is free software; you can redistribute it and/or
6
 * modify it under the terms of the GNU Lesser General Public
7
 * License as published by the Free Software Foundation; either
8
 * version 2 of the License, or (at your option) any later version.
9
 *
10
 * This library is distributed in the hope that it will be useful,
11
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13
 * Lesser General Public License for more details.
14
 *
15
 * You should have received a copy of the GNU Lesser General Public
16
 * License along with this library; if not, write to the Free Software
17
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
18
 *
19
 * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>
20
 */
21
#include "avcodec.h"
22
#include "dsputil.h"
23
#include "mpegvideo.h"
24

    
25
int ff_bit_exact=0;
26

    
27
UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];
28
UINT32 squareTbl[512];
29

    
30
const UINT8 ff_zigzag_direct[64] = {
31
    0,   1,  8, 16,  9,  2,  3, 10,
32
    17, 24, 32, 25, 18, 11,  4,  5,
33
    12, 19, 26, 33, 40, 48, 41, 34,
34
    27, 20, 13,  6,  7, 14, 21, 28,
35
    35, 42, 49, 56, 57, 50, 43, 36,
36
    29, 22, 15, 23, 30, 37, 44, 51,
37
    58, 59, 52, 45, 38, 31, 39, 46,
38
    53, 60, 61, 54, 47, 55, 62, 63
39
};
40

    
41
/* not permutated inverse zigzag_direct + 1 for MMX quantizer */
42
UINT16 __align8 inv_zigzag_direct16[64];
43

    
44
const UINT8 ff_alternate_horizontal_scan[64] = {
45
    0,  1,   2,  3,  8,  9, 16, 17, 
46
    10, 11,  4,  5,  6,  7, 15, 14,
47
    13, 12, 19, 18, 24, 25, 32, 33, 
48
    26, 27, 20, 21, 22, 23, 28, 29,
49
    30, 31, 34, 35, 40, 41, 48, 49, 
50
    42, 43, 36, 37, 38, 39, 44, 45,
51
    46, 47, 50, 51, 56, 57, 58, 59, 
52
    52, 53, 54, 55, 60, 61, 62, 63,
53
};
54

    
55
const UINT8 ff_alternate_vertical_scan[64] = {
56
    0,  8,  16, 24,  1,  9,  2, 10, 
57
    17, 25, 32, 40, 48, 56, 57, 49,
58
    41, 33, 26, 18,  3, 11,  4, 12, 
59
    19, 27, 34, 42, 50, 58, 35, 43,
60
    51, 59, 20, 28,  5, 13,  6, 14, 
61
    21, 29, 36, 44, 52, 60, 37, 45,
62
    53, 61, 22, 30,  7, 15, 23, 31, 
63
    38, 46, 54, 62, 39, 47, 55, 63,
64
};
65

    
66
/* a*inverse[b]>>32 == a/b for all 0<=a<=65536 && 2<=b<=255 */
67
const UINT32 inverse[256]={
68
         0, 4294967295U,2147483648U,1431655766, 1073741824,  858993460,  715827883,  613566757, 
69
 536870912,  477218589,  429496730,  390451573,  357913942,  330382100,  306783379,  286331154, 
70
 268435456,  252645136,  238609295,  226050911,  214748365,  204522253,  195225787,  186737709, 
71
 178956971,  171798692,  165191050,  159072863,  153391690,  148102321,  143165577,  138547333, 
72
 134217728,  130150525,  126322568,  122713352,  119304648,  116080198,  113025456,  110127367, 
73
 107374183,  104755300,  102261127,   99882961,   97612894,   95443718,   93368855,   91382283, 
74
  89478486,   87652394,   85899346,   84215046,   82595525,   81037119,   79536432,   78090315, 
75
  76695845,   75350304,   74051161,   72796056,   71582789,   70409300,   69273667,   68174085, 
76
  67108864,   66076420,   65075263,   64103990,   63161284,   62245903,   61356676,   60492498, 
77
  59652324,   58835169,   58040099,   57266231,   56512728,   55778797,   55063684,   54366675, 
78
  53687092,   53024288,   52377650,   51746594,   51130564,   50529028,   49941481,   49367441, 
79
  48806447,   48258060,   47721859,   47197443,   46684428,   46182445,   45691142,   45210183, 
80
  44739243,   44278014,   43826197,   43383509,   42949673,   42524429,   42107523,   41698712, 
81
  41297763,   40904451,   40518560,   40139882,   39768216,   39403370,   39045158,   38693400, 
82
  38347923,   38008561,   37675152,   37347542,   37025581,   36709123,   36398028,   36092163, 
83
  35791395,   35495598,   35204650,   34918434,   34636834,   34359739,   34087043,   33818641, 
84
  33554432,   33294321,   33038210,   32786010,   32537632,   32292988,   32051995,   31814573, 
85
  31580642,   31350127,   31122952,   30899046,   30678338,   30460761,   30246249,   30034737, 
86
  29826162,   29620465,   29417585,   29217465,   29020050,   28825284,   28633116,   28443493, 
87
  28256364,   28071682,   27889399,   27709467,   27531842,   27356480,   27183338,   27012373, 
88
  26843546,   26676816,   26512144,   26349493,   26188825,   26030105,   25873297,   25718368, 
89
  25565282,   25414008,   25264514,   25116768,   24970741,   24826401,   24683721,   24542671, 
90
  24403224,   24265352,   24129030,   23994231,   23860930,   23729102,   23598722,   23469767, 
91
  23342214,   23216040,   23091223,   22967740,   22845571,   22724695,   22605092,   22486740, 
92
  22369622,   22253717,   22139007,   22025474,   21913099,   21801865,   21691755,   21582751, 
93
  21474837,   21367997,   21262215,   21157475,   21053762,   20951060,   20849356,   20748635, 
94
  20648882,   20550083,   20452226,   20355296,   20259280,   20164166,   20069941,   19976593, 
95
  19884108,   19792477,   19701685,   19611723,   19522579,   19434242,   19346700,   19259944, 
96
  19173962,   19088744,   19004281,   18920561,   18837576,   18755316,   18673771,   18592933, 
97
  18512791,   18433337,   18354562,   18276457,   18199014,   18122225,   18046082,   17970575, 
98
  17895698,   17821442,   17747799,   17674763,   17602325,   17530479,   17459217,   17388532, 
99
  17318417,   17248865,   17179870,   17111424,   17043522,   16976156,   16909321,   16843010,
100
};
101

    
102
static int pix_sum_c(UINT8 * pix, int line_size)
103
{
104
    int s, i, j;
105

    
106
    s = 0;
107
    for (i = 0; i < 16; i++) {
108
        for (j = 0; j < 16; j += 8) {
109
            s += pix[0];
110
            s += pix[1];
111
            s += pix[2];
112
            s += pix[3];
113
            s += pix[4];
114
            s += pix[5];
115
            s += pix[6];
116
            s += pix[7];
117
            pix += 8;
118
        }
119
        pix += line_size - 16;
120
    }
121
    return s;
122
}
123

    
124
static int pix_norm1_c(UINT8 * pix, int line_size)
125
{
126
    int s, i, j;
127
    UINT32 *sq = squareTbl + 256;
128

    
129
    s = 0;
130
    for (i = 0; i < 16; i++) {
131
        for (j = 0; j < 16; j += 8) {
132
#if 0
133
            s += sq[pix[0]];
134
            s += sq[pix[1]];
135
            s += sq[pix[2]];
136
            s += sq[pix[3]];
137
            s += sq[pix[4]];
138
            s += sq[pix[5]];
139
            s += sq[pix[6]];
140
            s += sq[pix[7]];
141
#else
142
#if LONG_MAX > 2147483647
143
            register uint64_t x=*(uint64_t*)pix;
144
            s += sq[x&0xff];
145
            s += sq[(x>>8)&0xff];
146
            s += sq[(x>>16)&0xff];
147
            s += sq[(x>>24)&0xff];
148
            s += sq[(x>>32)&0xff];
149
            s += sq[(x>>40)&0xff];
150
            s += sq[(x>>48)&0xff];
151
            s += sq[(x>>56)&0xff];
152
#else
153
            register uint32_t x=*(uint32_t*)pix;
154
            s += sq[x&0xff];
155
            s += sq[(x>>8)&0xff];
156
            s += sq[(x>>16)&0xff];
157
            s += sq[(x>>24)&0xff];
158
            x=*(uint32_t*)(pix+4);
159
            s += sq[x&0xff];
160
            s += sq[(x>>8)&0xff];
161
            s += sq[(x>>16)&0xff];
162
            s += sq[(x>>24)&0xff];
163
#endif
164
#endif
165
            pix += 8;
166
        }
167
        pix += line_size - 16;
168
    }
169
    return s;
170
}
171

    
172

    
173
static int sse8_c(void *v, UINT8 * pix1, UINT8 * pix2, int line_size)
174
{
175
    int s, i;
176
    UINT32 *sq = squareTbl + 256;
177

    
178
    s = 0;
179
    for (i = 0; i < 8; i++) {
180
        s += sq[pix1[0] - pix2[0]];
181
        s += sq[pix1[1] - pix2[1]];
182
        s += sq[pix1[2] - pix2[2]];
183
        s += sq[pix1[3] - pix2[3]];
184
        s += sq[pix1[4] - pix2[4]];
185
        s += sq[pix1[5] - pix2[5]];
186
        s += sq[pix1[6] - pix2[6]];
187
        s += sq[pix1[7] - pix2[7]];
188
        pix1 += line_size;
189
        pix2 += line_size;
190
    }
191
    return s;
192
}
193

    
194
static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size)
195
{
196
    int s, i;
197
    uint32_t *sq = squareTbl + 256;
198

    
199
    s = 0;
200
    for (i = 0; i < 16; i++) {
201
        s += sq[pix1[ 0] - pix2[ 0]];
202
        s += sq[pix1[ 1] - pix2[ 1]];
203
        s += sq[pix1[ 2] - pix2[ 2]];
204
        s += sq[pix1[ 3] - pix2[ 3]];
205
        s += sq[pix1[ 4] - pix2[ 4]];
206
        s += sq[pix1[ 5] - pix2[ 5]];
207
        s += sq[pix1[ 6] - pix2[ 6]];
208
        s += sq[pix1[ 7] - pix2[ 7]];
209
        s += sq[pix1[ 8] - pix2[ 8]];
210
        s += sq[pix1[ 9] - pix2[ 9]];
211
        s += sq[pix1[10] - pix2[10]];
212
        s += sq[pix1[11] - pix2[11]];
213
        s += sq[pix1[12] - pix2[12]];
214
        s += sq[pix1[13] - pix2[13]];
215
        s += sq[pix1[14] - pix2[14]];
216
        s += sq[pix1[15] - pix2[15]];
217

    
218
        pix1 += line_size;
219
        pix2 += line_size;
220
    }
221
    return s;
222
}
223

    
224
static void get_pixels_c(DCTELEM *restrict block, const UINT8 *pixels, int line_size)
225
{
226
    int i;
227

    
228
    /* read the pixels */
229
    for(i=0;i<8;i++) {
230
        block[0] = pixels[0];
231
        block[1] = pixels[1];
232
        block[2] = pixels[2];
233
        block[3] = pixels[3];
234
        block[4] = pixels[4];
235
        block[5] = pixels[5];
236
        block[6] = pixels[6];
237
        block[7] = pixels[7];
238
        pixels += line_size;
239
        block += 8;
240
    }
241
}
242

    
243
static void diff_pixels_c(DCTELEM *restrict block, const UINT8 *s1,
244
                          const UINT8 *s2, int stride){
245
    int i;
246

    
247
    /* read the pixels */
248
    for(i=0;i<8;i++) {
249
        block[0] = s1[0] - s2[0];
250
        block[1] = s1[1] - s2[1];
251
        block[2] = s1[2] - s2[2];
252
        block[3] = s1[3] - s2[3];
253
        block[4] = s1[4] - s2[4];
254
        block[5] = s1[5] - s2[5];
255
        block[6] = s1[6] - s2[6];
256
        block[7] = s1[7] - s2[7];
257
        s1 += stride;
258
        s2 += stride;
259
        block += 8;
260
    }
261
}
262

    
263

    
264
static void put_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels,
265
                                 int line_size)
266
{
267
    int i;
268
    UINT8 *cm = cropTbl + MAX_NEG_CROP;
269
    
270
    /* read the pixels */
271
    for(i=0;i<8;i++) {
272
        pixels[0] = cm[block[0]];
273
        pixels[1] = cm[block[1]];
274
        pixels[2] = cm[block[2]];
275
        pixels[3] = cm[block[3]];
276
        pixels[4] = cm[block[4]];
277
        pixels[5] = cm[block[5]];
278
        pixels[6] = cm[block[6]];
279
        pixels[7] = cm[block[7]];
280

    
281
        pixels += line_size;
282
        block += 8;
283
    }
284
}
285

    
286
static void add_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels,
287
                          int line_size)
288
{
289
    int i;
290
    UINT8 *cm = cropTbl + MAX_NEG_CROP;
291
    
292
    /* read the pixels */
293
    for(i=0;i<8;i++) {
294
        pixels[0] = cm[pixels[0] + block[0]];
295
        pixels[1] = cm[pixels[1] + block[1]];
296
        pixels[2] = cm[pixels[2] + block[2]];
297
        pixels[3] = cm[pixels[3] + block[3]];
298
        pixels[4] = cm[pixels[4] + block[4]];
299
        pixels[5] = cm[pixels[5] + block[5]];
300
        pixels[6] = cm[pixels[6] + block[6]];
301
        pixels[7] = cm[pixels[7] + block[7]];
302
        pixels += line_size;
303
        block += 8;
304
    }
305
}
306
#if 0
307

308
#define PIXOP2(OPNAME, OP) \
309
static void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
310
{\
311
    int i;\
312
    for(i=0; i<h; i++){\
313
        OP(*((uint64_t*)block), LD64(pixels));\
314
        pixels+=line_size;\
315
        block +=line_size;\
316
    }\
317
}\
318
\
319
static void OPNAME ## _no_rnd_pixels_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
320
{\
321
    int i;\
322
    for(i=0; i<h; i++){\
323
        const uint64_t a= LD64(pixels  );\
324
        const uint64_t b= LD64(pixels+1);\
325
        OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
326
        pixels+=line_size;\
327
        block +=line_size;\
328
    }\
329
}\
330
\
331
static void OPNAME ## _pixels_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
332
{\
333
    int i;\
334
    for(i=0; i<h; i++){\
335
        const uint64_t a= LD64(pixels  );\
336
        const uint64_t b= LD64(pixels+1);\
337
        OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
338
        pixels+=line_size;\
339
        block +=line_size;\
340
    }\
341
}\
342
\
343
static void OPNAME ## _no_rnd_pixels_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
344
{\
345
    int i;\
346
    for(i=0; i<h; i++){\
347
        const uint64_t a= LD64(pixels          );\
348
        const uint64_t b= LD64(pixels+line_size);\
349
        OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
350
        pixels+=line_size;\
351
        block +=line_size;\
352
    }\
353
}\
354
\
355
static void OPNAME ## _pixels_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
356
{\
357
    int i;\
358
    for(i=0; i<h; i++){\
359
        const uint64_t a= LD64(pixels          );\
360
        const uint64_t b= LD64(pixels+line_size);\
361
        OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
362
        pixels+=line_size;\
363
        block +=line_size;\
364
    }\
365
}\
366
\
367
static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
368
{\
369
        int i;\
370
        const uint64_t a= LD64(pixels  );\
371
        const uint64_t b= LD64(pixels+1);\
372
        uint64_t l0=  (a&0x0303030303030303ULL)\
373
                    + (b&0x0303030303030303ULL)\
374
                    + 0x0202020202020202ULL;\
375
        uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
376
                   + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
377
        uint64_t l1,h1;\
378
\
379
        pixels+=line_size;\
380
        for(i=0; i<h; i+=2){\
381
            uint64_t a= LD64(pixels  );\
382
            uint64_t b= LD64(pixels+1);\
383
            l1=  (a&0x0303030303030303ULL)\
384
               + (b&0x0303030303030303ULL);\
385
            h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
386
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
387
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
388
            pixels+=line_size;\
389
            block +=line_size;\
390
            a= LD64(pixels  );\
391
            b= LD64(pixels+1);\
392
            l0=  (a&0x0303030303030303ULL)\
393
               + (b&0x0303030303030303ULL)\
394
               + 0x0202020202020202ULL;\
395
            h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
396
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
397
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
398
            pixels+=line_size;\
399
            block +=line_size;\
400
        }\
401
}\
402
\
403
static void OPNAME ## _no_rnd_pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
404
{\
405
        int i;\
406
        const uint64_t a= LD64(pixels  );\
407
        const uint64_t b= LD64(pixels+1);\
408
        uint64_t l0=  (a&0x0303030303030303ULL)\
409
                    + (b&0x0303030303030303ULL)\
410
                    + 0x0101010101010101ULL;\
411
        uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
412
                   + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
413
        uint64_t l1,h1;\
414
\
415
        pixels+=line_size;\
416
        for(i=0; i<h; i+=2){\
417
            uint64_t a= LD64(pixels  );\
418
            uint64_t b= LD64(pixels+1);\
419
            l1=  (a&0x0303030303030303ULL)\
420
               + (b&0x0303030303030303ULL);\
421
            h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
422
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
423
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
424
            pixels+=line_size;\
425
            block +=line_size;\
426
            a= LD64(pixels  );\
427
            b= LD64(pixels+1);\
428
            l0=  (a&0x0303030303030303ULL)\
429
               + (b&0x0303030303030303ULL)\
430
               + 0x0101010101010101ULL;\
431
            h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
432
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
433
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
434
            pixels+=line_size;\
435
            block +=line_size;\
436
        }\
437
}\
438
\
439
CALL_2X_PIXELS(OPNAME ## _pixels16_c    , OPNAME ## _pixels_c    , 8)\
440
CALL_2X_PIXELS(OPNAME ## _pixels16_x2_c , OPNAME ## _pixels_x2_c , 8)\
441
CALL_2X_PIXELS(OPNAME ## _pixels16_y2_c , OPNAME ## _pixels_y2_c , 8)\
442
CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_c, OPNAME ## _pixels_xy2_c, 8)\
443
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2_c , OPNAME ## _no_rnd_pixels_x2_c , 8)\
444
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2_c , OPNAME ## _no_rnd_pixels_y2_c , 8)\
445
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels_xy2_c, 8)
446

447
#define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEFEFEFEFEULL)>>1) )
448
#else // 64 bit variant
449

    
450
#define PIXOP2(OPNAME, OP) \
451
static void OPNAME ## _pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
452
    int i;\
453
    for(i=0; i<h; i++){\
454
        OP(*((uint32_t*)(block  )), LD32(pixels  ));\
455
        OP(*((uint32_t*)(block+4)), LD32(pixels+4));\
456
        pixels+=line_size;\
457
        block +=line_size;\
458
    }\
459
}\
460
static inline void OPNAME ## _no_rnd_pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
461
    OPNAME ## _pixels8_c(block, pixels, line_size, h);\
462
}\
463
\
464
static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
465
                                                int src_stride1, int src_stride2, int h){\
466
    int i;\
467
    for(i=0; i<h; i++){\
468
        uint32_t a,b;\
469
        a= LD32(&src1[i*src_stride1  ]);\
470
        b= LD32(&src2[i*src_stride2  ]);\
471
        OP(*((uint32_t*)&dst[i*dst_stride  ]), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\
472
        a= LD32(&src1[i*src_stride1+4]);\
473
        b= LD32(&src2[i*src_stride2+4]);\
474
        OP(*((uint32_t*)&dst[i*dst_stride+4]), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\
475
    }\
476
}\
477
\
478
static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
479
                                                int src_stride1, int src_stride2, int h){\
480
    int i;\
481
    for(i=0; i<h; i++){\
482
        uint32_t a,b;\
483
        a= LD32(&src1[i*src_stride1  ]);\
484
        b= LD32(&src2[i*src_stride2  ]);\
485
        OP(*((uint32_t*)&dst[i*dst_stride  ]), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\
486
        a= LD32(&src1[i*src_stride1+4]);\
487
        b= LD32(&src2[i*src_stride2+4]);\
488
        OP(*((uint32_t*)&dst[i*dst_stride+4]), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\
489
    }\
490
}\
491
\
492
static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
493
                                                int src_stride1, int src_stride2, int h){\
494
    OPNAME ## _pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\
495
    OPNAME ## _pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\
496
}\
497
\
498
static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
499
                                                int src_stride1, int src_stride2, int h){\
500
    OPNAME ## _no_rnd_pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\
501
    OPNAME ## _no_rnd_pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\
502
}\
503
\
504
static inline void OPNAME ## _no_rnd_pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
505
    OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
506
}\
507
\
508
static inline void OPNAME ## _pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
509
    OPNAME ## _pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
510
}\
511
\
512
static inline void OPNAME ## _no_rnd_pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
513
    OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
514
}\
515
\
516
static inline void OPNAME ## _pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
517
    OPNAME ## _pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
518
}\
519
\
520
static inline void OPNAME ## _pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
521
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
522
    int i;\
523
    for(i=0; i<h; i++){\
524
        uint32_t a, b, c, d, l0, l1, h0, h1;\
525
        a= LD32(&src1[i*src_stride1]);\
526
        b= LD32(&src2[i*src_stride2]);\
527
        c= LD32(&src3[i*src_stride3]);\
528
        d= LD32(&src4[i*src_stride4]);\
529
        l0=  (a&0x03030303UL)\
530
           + (b&0x03030303UL)\
531
           + 0x02020202UL;\
532
        h0= ((a&0xFCFCFCFCUL)>>2)\
533
          + ((b&0xFCFCFCFCUL)>>2);\
534
        l1=  (c&0x03030303UL)\
535
           + (d&0x03030303UL);\
536
        h1= ((c&0xFCFCFCFCUL)>>2)\
537
          + ((d&0xFCFCFCFCUL)>>2);\
538
        OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
539
        a= LD32(&src1[i*src_stride1+4]);\
540
        b= LD32(&src2[i*src_stride2+4]);\
541
        c= LD32(&src3[i*src_stride3+4]);\
542
        d= LD32(&src4[i*src_stride4+4]);\
543
        l0=  (a&0x03030303UL)\
544
           + (b&0x03030303UL)\
545
           + 0x02020202UL;\
546
        h0= ((a&0xFCFCFCFCUL)>>2)\
547
          + ((b&0xFCFCFCFCUL)>>2);\
548
        l1=  (c&0x03030303UL)\
549
           + (d&0x03030303UL);\
550
        h1= ((c&0xFCFCFCFCUL)>>2)\
551
          + ((d&0xFCFCFCFCUL)>>2);\
552
        OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
553
    }\
554
}\
555
static inline void OPNAME ## _no_rnd_pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
556
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
557
    int i;\
558
    for(i=0; i<h; i++){\
559
        uint32_t a, b, c, d, l0, l1, h0, h1;\
560
        a= LD32(&src1[i*src_stride1]);\
561
        b= LD32(&src2[i*src_stride2]);\
562
        c= LD32(&src3[i*src_stride3]);\
563
        d= LD32(&src4[i*src_stride4]);\
564
        l0=  (a&0x03030303UL)\
565
           + (b&0x03030303UL)\
566
           + 0x01010101UL;\
567
        h0= ((a&0xFCFCFCFCUL)>>2)\
568
          + ((b&0xFCFCFCFCUL)>>2);\
569
        l1=  (c&0x03030303UL)\
570
           + (d&0x03030303UL);\
571
        h1= ((c&0xFCFCFCFCUL)>>2)\
572
          + ((d&0xFCFCFCFCUL)>>2);\
573
        OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
574
        a= LD32(&src1[i*src_stride1+4]);\
575
        b= LD32(&src2[i*src_stride2+4]);\
576
        c= LD32(&src3[i*src_stride3+4]);\
577
        d= LD32(&src4[i*src_stride4+4]);\
578
        l0=  (a&0x03030303UL)\
579
           + (b&0x03030303UL)\
580
           + 0x01010101UL;\
581
        h0= ((a&0xFCFCFCFCUL)>>2)\
582
          + ((b&0xFCFCFCFCUL)>>2);\
583
        l1=  (c&0x03030303UL)\
584
           + (d&0x03030303UL);\
585
        h1= ((c&0xFCFCFCFCUL)>>2)\
586
          + ((d&0xFCFCFCFCUL)>>2);\
587
        OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
588
    }\
589
}\
590
static inline void OPNAME ## _pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
591
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
592
    OPNAME ## _pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
593
    OPNAME ## _pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
594
}\
595
static inline void OPNAME ## _no_rnd_pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
596
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
597
    OPNAME ## _no_rnd_pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
598
    OPNAME ## _no_rnd_pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
599
}\
600
\
601
static inline void OPNAME ## _pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
602
{\
603
    int j;\
604
    for(j=0; j<2; j++){\
605
        int i;\
606
        const uint32_t a= LD32(pixels  );\
607
        const uint32_t b= LD32(pixels+1);\
608
        uint32_t l0=  (a&0x03030303UL)\
609
                    + (b&0x03030303UL)\
610
                    + 0x02020202UL;\
611
        uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
612
                   + ((b&0xFCFCFCFCUL)>>2);\
613
        uint32_t l1,h1;\
614
\
615
        pixels+=line_size;\
616
        for(i=0; i<h; i+=2){\
617
            uint32_t a= LD32(pixels  );\
618
            uint32_t b= LD32(pixels+1);\
619
            l1=  (a&0x03030303UL)\
620
               + (b&0x03030303UL);\
621
            h1= ((a&0xFCFCFCFCUL)>>2)\
622
              + ((b&0xFCFCFCFCUL)>>2);\
623
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
624
            pixels+=line_size;\
625
            block +=line_size;\
626
            a= LD32(pixels  );\
627
            b= LD32(pixels+1);\
628
            l0=  (a&0x03030303UL)\
629
               + (b&0x03030303UL)\
630
               + 0x02020202UL;\
631
            h0= ((a&0xFCFCFCFCUL)>>2)\
632
              + ((b&0xFCFCFCFCUL)>>2);\
633
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
634
            pixels+=line_size;\
635
            block +=line_size;\
636
        }\
637
        pixels+=4-line_size*(h+1);\
638
        block +=4-line_size*h;\
639
    }\
640
}\
641
\
642
static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
643
{\
644
    int j;\
645
    for(j=0; j<2; j++){\
646
        int i;\
647
        const uint32_t a= LD32(pixels  );\
648
        const uint32_t b= LD32(pixels+1);\
649
        uint32_t l0=  (a&0x03030303UL)\
650
                    + (b&0x03030303UL)\
651
                    + 0x01010101UL;\
652
        uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
653
                   + ((b&0xFCFCFCFCUL)>>2);\
654
        uint32_t l1,h1;\
655
\
656
        pixels+=line_size;\
657
        for(i=0; i<h; i+=2){\
658
            uint32_t a= LD32(pixels  );\
659
            uint32_t b= LD32(pixels+1);\
660
            l1=  (a&0x03030303UL)\
661
               + (b&0x03030303UL);\
662
            h1= ((a&0xFCFCFCFCUL)>>2)\
663
              + ((b&0xFCFCFCFCUL)>>2);\
664
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
665
            pixels+=line_size;\
666
            block +=line_size;\
667
            a= LD32(pixels  );\
668
            b= LD32(pixels+1);\
669
            l0=  (a&0x03030303UL)\
670
               + (b&0x03030303UL)\
671
               + 0x01010101UL;\
672
            h0= ((a&0xFCFCFCFCUL)>>2)\
673
              + ((b&0xFCFCFCFCUL)>>2);\
674
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
675
            pixels+=line_size;\
676
            block +=line_size;\
677
        }\
678
        pixels+=4-line_size*(h+1);\
679
        block +=4-line_size*h;\
680
    }\
681
}\
682
\
683
CALL_2X_PIXELS(OPNAME ## _pixels16_c  , OPNAME ## _pixels8_c  , 8)\
684
CALL_2X_PIXELS(OPNAME ## _pixels16_x2_c , OPNAME ## _pixels8_x2_c , 8)\
685
CALL_2X_PIXELS(OPNAME ## _pixels16_y2_c , OPNAME ## _pixels8_y2_c , 8)\
686
CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_c, OPNAME ## _pixels8_xy2_c, 8)\
687
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_c  , OPNAME ## _pixels8_c         , 8)\
688
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2_c , OPNAME ## _no_rnd_pixels8_x2_c , 8)\
689
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2_c , OPNAME ## _no_rnd_pixels8_y2_c , 8)\
690
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels8_xy2_c, 8)\
691

    
692
#define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEUL)>>1) )
693
#endif
694
#define op_put(a, b) a = b
695

    
696
PIXOP2(avg, op_avg)
697
PIXOP2(put, op_put)
698
#undef op_avg
699
#undef op_put
700

    
701
#define avg2(a,b) ((a+b+1)>>1)
702
#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
703

    
704

    
705
static void gmc1_c(UINT8 *dst, UINT8 *src, int stride, int h, int x16, int y16, int rounder)
706
{
707
    const int A=(16-x16)*(16-y16);
708
    const int B=(   x16)*(16-y16);
709
    const int C=(16-x16)*(   y16);
710
    const int D=(   x16)*(   y16);
711
    int i;
712

    
713
    for(i=0; i<h; i++)
714
    {
715
        dst[0]= (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + rounder)>>8;
716
        dst[1]= (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + rounder)>>8;
717
        dst[2]= (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + rounder)>>8;
718
        dst[3]= (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + rounder)>>8;
719
        dst[4]= (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + rounder)>>8;
720
        dst[5]= (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + rounder)>>8;
721
        dst[6]= (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + rounder)>>8;
722
        dst[7]= (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + rounder)>>8;
723
        dst+= stride;
724
        src+= stride;
725
    }
726
}
727

    
728
static void gmc_c(UINT8 *dst, UINT8 *src, int stride, int h, int ox, int oy, 
729
                  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
730
{
731
    int y, vx, vy;
732
    const int s= 1<<shift;
733
    
734
    width--;
735
    height--;
736

    
737
    for(y=0; y<h; y++){
738
        int x;
739

    
740
        vx= ox;
741
        vy= oy;
742
        for(x=0; x<8; x++){ //XXX FIXME optimize
743
            int src_x, src_y, frac_x, frac_y, index;
744

    
745
            src_x= vx>>16;
746
            src_y= vy>>16;
747
            frac_x= src_x&(s-1);
748
            frac_y= src_y&(s-1);
749
            src_x>>=shift;
750
            src_y>>=shift;
751
  
752
            if((unsigned)src_x < width){
753
                if((unsigned)src_y < height){
754
                    index= src_x + src_y*stride;
755
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_x)
756
                                           + src[index       +1]*   frac_x )*(s-frac_y)
757
                                        + (  src[index+stride  ]*(s-frac_x)
758
                                           + src[index+stride+1]*   frac_x )*   frac_y
759
                                        + r)>>(shift*2);
760
                }else{
761
                    index= src_x + clip(src_y, 0, height)*stride;                    
762
                    dst[y*stride + x]= ( (  src[index         ]*(s-frac_x) 
763
                                          + src[index       +1]*   frac_x )*s
764
                                        + r)>>(shift*2);
765
                }
766
            }else{
767
                if((unsigned)src_y < height){
768
                    index= clip(src_x, 0, width) + src_y*stride;                    
769
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_y) 
770
                                           + src[index+stride  ]*   frac_y )*s
771
                                        + r)>>(shift*2);
772
                }else{
773
                    index= clip(src_x, 0, width) + clip(src_y, 0, height)*stride;                    
774
                    dst[y*stride + x]=    src[index         ];
775
                }
776
            }
777
            
778
            vx+= dxx;
779
            vy+= dyx;
780
        }
781
        ox += dxy;
782
        oy += dyy;
783
    }
784
}
785

    
786
static inline void copy_block17(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h)
787
{
788
    int i;
789
    for(i=0; i<h; i++)
790
    {
791
        ST32(dst   , LD32(src   ));
792
        ST32(dst+4 , LD32(src+4 ));
793
        ST32(dst+8 , LD32(src+8 ));
794
        ST32(dst+12, LD32(src+12));
795
        dst[16]= src[16];
796
        dst+=dstStride;
797
        src+=srcStride;
798
    }
799
}
800

    
801
static inline void copy_block9(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h)
802
{
803
    int i;
804
    for(i=0; i<h; i++)
805
    {
806
        ST32(dst   , LD32(src   ));
807
        ST32(dst+4 , LD32(src+4 ));
808
        dst[8]= src[8];
809
        dst+=dstStride;
810
        src+=srcStride;
811
    }
812
}
813

    
814

    
815
#define QPEL_MC(r, OPNAME, RND, OP) \
816
static void OPNAME ## mpeg4_qpel8_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h){\
817
    UINT8 *cm = cropTbl + MAX_NEG_CROP;\
818
    int i;\
819
    for(i=0; i<h; i++)\
820
    {\
821
        OP(dst[0], (src[0]+src[1])*20 - (src[0]+src[2])*6 + (src[1]+src[3])*3 - (src[2]+src[4]));\
822
        OP(dst[1], (src[1]+src[2])*20 - (src[0]+src[3])*6 + (src[0]+src[4])*3 - (src[1]+src[5]));\
823
        OP(dst[2], (src[2]+src[3])*20 - (src[1]+src[4])*6 + (src[0]+src[5])*3 - (src[0]+src[6]));\
824
        OP(dst[3], (src[3]+src[4])*20 - (src[2]+src[5])*6 + (src[1]+src[6])*3 - (src[0]+src[7]));\
825
        OP(dst[4], (src[4]+src[5])*20 - (src[3]+src[6])*6 + (src[2]+src[7])*3 - (src[1]+src[8]));\
826
        OP(dst[5], (src[5]+src[6])*20 - (src[4]+src[7])*6 + (src[3]+src[8])*3 - (src[2]+src[8]));\
827
        OP(dst[6], (src[6]+src[7])*20 - (src[5]+src[8])*6 + (src[4]+src[8])*3 - (src[3]+src[7]));\
828
        OP(dst[7], (src[7]+src[8])*20 - (src[6]+src[8])*6 + (src[5]+src[7])*3 - (src[4]+src[6]));\
829
        dst+=dstStride;\
830
        src+=srcStride;\
831
    }\
832
}\
833
\
834
static void OPNAME ## mpeg4_qpel8_v_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride){\
835
    const int w=8;\
836
    UINT8 *cm = cropTbl + MAX_NEG_CROP;\
837
    int i;\
838
    for(i=0; i<w; i++)\
839
    {\
840
        const int src0= src[0*srcStride];\
841
        const int src1= src[1*srcStride];\
842
        const int src2= src[2*srcStride];\
843
        const int src3= src[3*srcStride];\
844
        const int src4= src[4*srcStride];\
845
        const int src5= src[5*srcStride];\
846
        const int src6= src[6*srcStride];\
847
        const int src7= src[7*srcStride];\
848
        const int src8= src[8*srcStride];\
849
        OP(dst[0*dstStride], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
850
        OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
851
        OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
852
        OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
853
        OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
854
        OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
855
        OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
856
        OP(dst[7*dstStride], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
857
        dst++;\
858
        src++;\
859
    }\
860
}\
861
\
862
static void OPNAME ## mpeg4_qpel16_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h){\
863
    UINT8 *cm = cropTbl + MAX_NEG_CROP;\
864
    int i;\
865
    \
866
    for(i=0; i<h; i++)\
867
    {\
868
        OP(dst[ 0], (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]));\
869
        OP(dst[ 1], (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]));\
870
        OP(dst[ 2], (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]));\
871
        OP(dst[ 3], (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]));\
872
        OP(dst[ 4], (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]));\
873
        OP(dst[ 5], (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 9]));\
874
        OP(dst[ 6], (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 9])*3 - (src[ 3]+src[10]));\
875
        OP(dst[ 7], (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 9])*6 + (src[ 5]+src[10])*3 - (src[ 4]+src[11]));\
876
        OP(dst[ 8], (src[ 8]+src[ 9])*20 - (src[ 7]+src[10])*6 + (src[ 6]+src[11])*3 - (src[ 5]+src[12]));\
877
        OP(dst[ 9], (src[ 9]+src[10])*20 - (src[ 8]+src[11])*6 + (src[ 7]+src[12])*3 - (src[ 6]+src[13]));\
878
        OP(dst[10], (src[10]+src[11])*20 - (src[ 9]+src[12])*6 + (src[ 8]+src[13])*3 - (src[ 7]+src[14]));\
879
        OP(dst[11], (src[11]+src[12])*20 - (src[10]+src[13])*6 + (src[ 9]+src[14])*3 - (src[ 8]+src[15]));\
880
        OP(dst[12], (src[12]+src[13])*20 - (src[11]+src[14])*6 + (src[10]+src[15])*3 - (src[ 9]+src[16]));\
881
        OP(dst[13], (src[13]+src[14])*20 - (src[12]+src[15])*6 + (src[11]+src[16])*3 - (src[10]+src[16]));\
882
        OP(dst[14], (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]));\
883
        OP(dst[15], (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]));\
884
        dst+=dstStride;\
885
        src+=srcStride;\
886
    }\
887
}\
888
\
889
static void OPNAME ## mpeg4_qpel16_v_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride){\
890
    UINT8 *cm = cropTbl + MAX_NEG_CROP;\
891
    int i;\
892
    const int w=16;\
893
    for(i=0; i<w; i++)\
894
    {\
895
        const int src0= src[0*srcStride];\
896
        const int src1= src[1*srcStride];\
897
        const int src2= src[2*srcStride];\
898
        const int src3= src[3*srcStride];\
899
        const int src4= src[4*srcStride];\
900
        const int src5= src[5*srcStride];\
901
        const int src6= src[6*srcStride];\
902
        const int src7= src[7*srcStride];\
903
        const int src8= src[8*srcStride];\
904
        const int src9= src[9*srcStride];\
905
        const int src10= src[10*srcStride];\
906
        const int src11= src[11*srcStride];\
907
        const int src12= src[12*srcStride];\
908
        const int src13= src[13*srcStride];\
909
        const int src14= src[14*srcStride];\
910
        const int src15= src[15*srcStride];\
911
        const int src16= src[16*srcStride];\
912
        OP(dst[ 0*dstStride], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
913
        OP(dst[ 1*dstStride], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
914
        OP(dst[ 2*dstStride], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
915
        OP(dst[ 3*dstStride], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
916
        OP(dst[ 4*dstStride], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
917
        OP(dst[ 5*dstStride], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
918
        OP(dst[ 6*dstStride], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
919
        OP(dst[ 7*dstStride], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
920
        OP(dst[ 8*dstStride], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
921
        OP(dst[ 9*dstStride], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
922
        OP(dst[10*dstStride], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
923
        OP(dst[11*dstStride], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
924
        OP(dst[12*dstStride], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
925
        OP(dst[13*dstStride], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
926
        OP(dst[14*dstStride], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
927
        OP(dst[15*dstStride], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
928
        dst++;\
929
        src++;\
930
    }\
931
}\
932
\
933
static void OPNAME ## qpel8_mc00_c (UINT8 *dst, UINT8 *src, int stride){\
934
    OPNAME ## pixels8_c(dst, src, stride, 8);\
935
}\
936
\
937
static void OPNAME ## qpel8_mc10_c(UINT8 *dst, UINT8 *src, int stride){\
938
    UINT8 half[64];\
939
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
940
    OPNAME ## pixels8_l2(dst, src, half, stride, stride, 8, 8);\
941
}\
942
\
943
static void OPNAME ## qpel8_mc20_c(UINT8 *dst, UINT8 *src, int stride){\
944
    OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
945
}\
946
\
947
static void OPNAME ## qpel8_mc30_c(UINT8 *dst, UINT8 *src, int stride){\
948
    UINT8 half[64];\
949
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
950
    OPNAME ## pixels8_l2(dst, src+1, half, stride, stride, 8, 8);\
951
}\
952
\
953
static void OPNAME ## qpel8_mc01_c(UINT8 *dst, UINT8 *src, int stride){\
954
    UINT8 full[16*9];\
955
    UINT8 half[64];\
956
    copy_block9(full, src, 16, stride, 9);\
957
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
958
    OPNAME ## pixels8_l2(dst, full, half, stride, 16, 8, 8);\
959
}\
960
\
961
static void OPNAME ## qpel8_mc02_c(UINT8 *dst, UINT8 *src, int stride){\
962
    UINT8 full[16*9];\
963
    copy_block9(full, src, 16, stride, 9);\
964
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
965
}\
966
\
967
static void OPNAME ## qpel8_mc03_c(UINT8 *dst, UINT8 *src, int stride){\
968
    UINT8 full[16*9];\
969
    UINT8 half[64];\
970
    copy_block9(full, src, 16, stride, 9);\
971
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
972
    OPNAME ## pixels8_l2(dst, full+16, half, stride, 16, 8, 8);\
973
}\
974
void ff_ ## OPNAME ## qpel8_mc11_old_c(UINT8 *dst, UINT8 *src, int stride){\
975
    UINT8 full[16*9];\
976
    UINT8 halfH[72];\
977
    UINT8 halfV[64];\
978
    UINT8 halfHV[64];\
979
    copy_block9(full, src, 16, stride, 9);\
980
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
981
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
982
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
983
    OPNAME ## pixels8_l4(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
984
}\
985
static void OPNAME ## qpel8_mc11_c(UINT8 *dst, UINT8 *src, int stride){\
986
    UINT8 full[16*9];\
987
    UINT8 halfH[72];\
988
    UINT8 halfHV[64];\
989
    copy_block9(full, src, 16, stride, 9);\
990
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
991
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
992
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
993
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
994
}\
995
void ff_ ## OPNAME ## qpel8_mc31_old_c(UINT8 *dst, UINT8 *src, int stride){\
996
    UINT8 full[16*9];\
997
    UINT8 halfH[72];\
998
    UINT8 halfV[64];\
999
    UINT8 halfHV[64];\
1000
    copy_block9(full, src, 16, stride, 9);\
1001
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1002
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
1003
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1004
    OPNAME ## pixels8_l4(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
1005
}\
1006
static void OPNAME ## qpel8_mc31_c(UINT8 *dst, UINT8 *src, int stride){\
1007
    UINT8 full[16*9];\
1008
    UINT8 halfH[72];\
1009
    UINT8 halfHV[64];\
1010
    copy_block9(full, src, 16, stride, 9);\
1011
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1012
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
1013
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1014
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
1015
}\
1016
void ff_ ## OPNAME ## qpel8_mc13_old_c(UINT8 *dst, UINT8 *src, int stride){\
1017
    UINT8 full[16*9];\
1018
    UINT8 halfH[72];\
1019
    UINT8 halfV[64];\
1020
    UINT8 halfHV[64];\
1021
    copy_block9(full, src, 16, stride, 9);\
1022
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1023
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
1024
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1025
    OPNAME ## pixels8_l4(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
1026
}\
1027
static void OPNAME ## qpel8_mc13_c(UINT8 *dst, UINT8 *src, int stride){\
1028
    UINT8 full[16*9];\
1029
    UINT8 halfH[72];\
1030
    UINT8 halfHV[64];\
1031
    copy_block9(full, src, 16, stride, 9);\
1032
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1033
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
1034
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1035
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
1036
}\
1037
void ff_ ## OPNAME ## qpel8_mc33_old_c(UINT8 *dst, UINT8 *src, int stride){\
1038
    UINT8 full[16*9];\
1039
    UINT8 halfH[72];\
1040
    UINT8 halfV[64];\
1041
    UINT8 halfHV[64];\
1042
    copy_block9(full, src, 16, stride, 9);\
1043
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full  , 8, 16, 9);\
1044
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
1045
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1046
    OPNAME ## pixels8_l4(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
1047
}\
1048
static void OPNAME ## qpel8_mc33_c(UINT8 *dst, UINT8 *src, int stride){\
1049
    UINT8 full[16*9];\
1050
    UINT8 halfH[72];\
1051
    UINT8 halfHV[64];\
1052
    copy_block9(full, src, 16, stride, 9);\
1053
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1054
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
1055
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1056
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
1057
}\
1058
static void OPNAME ## qpel8_mc21_c(UINT8 *dst, UINT8 *src, int stride){\
1059
    UINT8 halfH[72];\
1060
    UINT8 halfHV[64];\
1061
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
1062
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1063
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
1064
}\
1065
static void OPNAME ## qpel8_mc23_c(UINT8 *dst, UINT8 *src, int stride){\
1066
    UINT8 halfH[72];\
1067
    UINT8 halfHV[64];\
1068
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
1069
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1070
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
1071
}\
1072
void ff_ ## OPNAME ## qpel8_mc12_old_c(UINT8 *dst, UINT8 *src, int stride){\
1073
    UINT8 full[16*9];\
1074
    UINT8 halfH[72];\
1075
    UINT8 halfV[64];\
1076
    UINT8 halfHV[64];\
1077
    copy_block9(full, src, 16, stride, 9);\
1078
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1079
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
1080
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1081
    OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\
1082
}\
1083
static void OPNAME ## qpel8_mc12_c(UINT8 *dst, UINT8 *src, int stride){\
1084
    UINT8 full[16*9];\
1085
    UINT8 halfH[72];\
1086
    copy_block9(full, src, 16, stride, 9);\
1087
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1088
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
1089
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1090
}\
1091
void ff_ ## OPNAME ## qpel8_mc32_old_c(UINT8 *dst, UINT8 *src, int stride){\
1092
    UINT8 full[16*9];\
1093
    UINT8 halfH[72];\
1094
    UINT8 halfV[64];\
1095
    UINT8 halfHV[64];\
1096
    copy_block9(full, src, 16, stride, 9);\
1097
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1098
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
1099
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
1100
    OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\
1101
}\
1102
static void OPNAME ## qpel8_mc32_c(UINT8 *dst, UINT8 *src, int stride){\
1103
    UINT8 full[16*9];\
1104
    UINT8 halfH[72];\
1105
    copy_block9(full, src, 16, stride, 9);\
1106
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
1107
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
1108
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1109
}\
1110
static void OPNAME ## qpel8_mc22_c(UINT8 *dst, UINT8 *src, int stride){\
1111
    UINT8 halfH[72];\
1112
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
1113
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
1114
}\
1115
static void OPNAME ## qpel16_mc00_c (UINT8 *dst, UINT8 *src, int stride){\
1116
    OPNAME ## pixels16_c(dst, src, stride, 16);\
1117
}\
1118
\
1119
static void OPNAME ## qpel16_mc10_c(UINT8 *dst, UINT8 *src, int stride){\
1120
    UINT8 half[256];\
1121
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
1122
    OPNAME ## pixels16_l2(dst, src, half, stride, stride, 16, 16);\
1123
}\
1124
\
1125
static void OPNAME ## qpel16_mc20_c(UINT8 *dst, UINT8 *src, int stride){\
1126
    OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
1127
}\
1128
\
1129
static void OPNAME ## qpel16_mc30_c(UINT8 *dst, UINT8 *src, int stride){\
1130
    UINT8 half[256];\
1131
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
1132
    OPNAME ## pixels16_l2(dst, src+1, half, stride, stride, 16, 16);\
1133
}\
1134
\
1135
static void OPNAME ## qpel16_mc01_c(UINT8 *dst, UINT8 *src, int stride){\
1136
    UINT8 full[24*17];\
1137
    UINT8 half[256];\
1138
    copy_block17(full, src, 24, stride, 17);\
1139
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
1140
    OPNAME ## pixels16_l2(dst, full, half, stride, 24, 16, 16);\
1141
}\
1142
\
1143
static void OPNAME ## qpel16_mc02_c(UINT8 *dst, UINT8 *src, int stride){\
1144
    UINT8 full[24*17];\
1145
    copy_block17(full, src, 24, stride, 17);\
1146
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
1147
}\
1148
\
1149
static void OPNAME ## qpel16_mc03_c(UINT8 *dst, UINT8 *src, int stride){\
1150
    UINT8 full[24*17];\
1151
    UINT8 half[256];\
1152
    copy_block17(full, src, 24, stride, 17);\
1153
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
1154
    OPNAME ## pixels16_l2(dst, full+24, half, stride, 24, 16, 16);\
1155
}\
1156
void ff_ ## OPNAME ## qpel16_mc11_old_c(UINT8 *dst, UINT8 *src, int stride){\
1157
    UINT8 full[24*17];\
1158
    UINT8 halfH[272];\
1159
    UINT8 halfV[256];\
1160
    UINT8 halfHV[256];\
1161
    copy_block17(full, src, 24, stride, 17);\
1162
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1163
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1164
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1165
    OPNAME ## pixels16_l4(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1166
}\
1167
static void OPNAME ## qpel16_mc11_c(UINT8 *dst, UINT8 *src, int stride){\
1168
    UINT8 full[24*17];\
1169
    UINT8 halfH[272];\
1170
    UINT8 halfHV[256];\
1171
    copy_block17(full, src, 24, stride, 17);\
1172
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1173
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
1174
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1175
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
1176
}\
1177
void ff_ ## OPNAME ## qpel16_mc31_old_c(UINT8 *dst, UINT8 *src, int stride){\
1178
    UINT8 full[24*17];\
1179
    UINT8 halfH[272];\
1180
    UINT8 halfV[256];\
1181
    UINT8 halfHV[256];\
1182
    copy_block17(full, src, 24, stride, 17);\
1183
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1184
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1185
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1186
    OPNAME ## pixels16_l4(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1187
}\
1188
static void OPNAME ## qpel16_mc31_c(UINT8 *dst, UINT8 *src, int stride){\
1189
    UINT8 full[24*17];\
1190
    UINT8 halfH[272];\
1191
    UINT8 halfHV[256];\
1192
    copy_block17(full, src, 24, stride, 17);\
1193
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1194
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
1195
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1196
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
1197
}\
1198
void ff_ ## OPNAME ## qpel16_mc13_old_c(UINT8 *dst, UINT8 *src, int stride){\
1199
    UINT8 full[24*17];\
1200
    UINT8 halfH[272];\
1201
    UINT8 halfV[256];\
1202
    UINT8 halfHV[256];\
1203
    copy_block17(full, src, 24, stride, 17);\
1204
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1205
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1206
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1207
    OPNAME ## pixels16_l4(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1208
}\
1209
static void OPNAME ## qpel16_mc13_c(UINT8 *dst, UINT8 *src, int stride){\
1210
    UINT8 full[24*17];\
1211
    UINT8 halfH[272];\
1212
    UINT8 halfHV[256];\
1213
    copy_block17(full, src, 24, stride, 17);\
1214
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1215
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
1216
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1217
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1218
}\
1219
void ff_ ## OPNAME ## qpel16_mc33_old_c(UINT8 *dst, UINT8 *src, int stride){\
1220
    UINT8 full[24*17];\
1221
    UINT8 halfH[272];\
1222
    UINT8 halfV[256];\
1223
    UINT8 halfHV[256];\
1224
    copy_block17(full, src, 24, stride, 17);\
1225
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\
1226
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1227
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1228
    OPNAME ## pixels16_l4(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1229
}\
1230
static void OPNAME ## qpel16_mc33_c(UINT8 *dst, UINT8 *src, int stride){\
1231
    UINT8 full[24*17];\
1232
    UINT8 halfH[272];\
1233
    UINT8 halfHV[256];\
1234
    copy_block17(full, src, 24, stride, 17);\
1235
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1236
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
1237
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1238
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1239
}\
1240
static void OPNAME ## qpel16_mc21_c(UINT8 *dst, UINT8 *src, int stride){\
1241
    UINT8 halfH[272];\
1242
    UINT8 halfHV[256];\
1243
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1244
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1245
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
1246
}\
1247
static void OPNAME ## qpel16_mc23_c(UINT8 *dst, UINT8 *src, int stride){\
1248
    UINT8 halfH[272];\
1249
    UINT8 halfHV[256];\
1250
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1251
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1252
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1253
}\
1254
void ff_ ## OPNAME ## qpel16_mc12_old_c(UINT8 *dst, UINT8 *src, int stride){\
1255
    UINT8 full[24*17];\
1256
    UINT8 halfH[272];\
1257
    UINT8 halfV[256];\
1258
    UINT8 halfHV[256];\
1259
    copy_block17(full, src, 24, stride, 17);\
1260
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1261
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1262
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1263
    OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\
1264
}\
1265
static void OPNAME ## qpel16_mc12_c(UINT8 *dst, UINT8 *src, int stride){\
1266
    UINT8 full[24*17];\
1267
    UINT8 halfH[272];\
1268
    copy_block17(full, src, 24, stride, 17);\
1269
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1270
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
1271
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1272
}\
1273
void ff_ ## OPNAME ## qpel16_mc32_old_c(UINT8 *dst, UINT8 *src, int stride){\
1274
    UINT8 full[24*17];\
1275
    UINT8 halfH[272];\
1276
    UINT8 halfV[256];\
1277
    UINT8 halfHV[256];\
1278
    copy_block17(full, src, 24, stride, 17);\
1279
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1280
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1281
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1282
    OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\
1283
}\
1284
static void OPNAME ## qpel16_mc32_c(UINT8 *dst, UINT8 *src, int stride){\
1285
    UINT8 full[24*17];\
1286
    UINT8 halfH[272];\
1287
    copy_block17(full, src, 24, stride, 17);\
1288
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1289
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
1290
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1291
}\
1292
static void OPNAME ## qpel16_mc22_c(UINT8 *dst, UINT8 *src, int stride){\
1293
    UINT8 halfH[272];\
1294
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1295
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1296
}
1297

    
1298
#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
1299
#define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
1300
#define op_put(a, b) a = cm[((b) + 16)>>5]
1301
#define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]
1302

    
1303
QPEL_MC(0, put_       , _       , op_put)
1304
QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
1305
QPEL_MC(0, avg_       , _       , op_avg)
1306
//QPEL_MC(1, avg_no_rnd , _       , op_avg)
1307
#undef op_avg
1308
#undef op_avg_no_rnd
1309
#undef op_put
1310
#undef op_put_no_rnd
1311

    
1312
static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
1313
    uint8_t *cm = cropTbl + MAX_NEG_CROP;
1314
    int i;
1315

    
1316
    for(i=0; i<h; i++){
1317
        dst[0]= cm[(9*(src[0] + src[1]) - (src[-1] + src[2]) + 8)>>4];
1318
        dst[1]= cm[(9*(src[1] + src[2]) - (src[ 0] + src[3]) + 8)>>4];
1319
        dst[2]= cm[(9*(src[2] + src[3]) - (src[ 1] + src[4]) + 8)>>4];
1320
        dst[3]= cm[(9*(src[3] + src[4]) - (src[ 2] + src[5]) + 8)>>4];
1321
        dst[4]= cm[(9*(src[4] + src[5]) - (src[ 3] + src[6]) + 8)>>4];
1322
        dst[5]= cm[(9*(src[5] + src[6]) - (src[ 4] + src[7]) + 8)>>4];
1323
        dst[6]= cm[(9*(src[6] + src[7]) - (src[ 5] + src[8]) + 8)>>4];
1324
        dst[7]= cm[(9*(src[7] + src[8]) - (src[ 6] + src[9]) + 8)>>4];
1325
        dst+=dstStride;
1326
        src+=srcStride;        
1327
    }
1328
}
1329

    
1330
static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
1331
    uint8_t *cm = cropTbl + MAX_NEG_CROP;
1332
    int i;
1333

    
1334
    for(i=0; i<w; i++){
1335
        const int src_1= src[ -srcStride];
1336
        const int src0 = src[0          ];
1337
        const int src1 = src[  srcStride];
1338
        const int src2 = src[2*srcStride];
1339
        const int src3 = src[3*srcStride];
1340
        const int src4 = src[4*srcStride];
1341
        const int src5 = src[5*srcStride];
1342
        const int src6 = src[6*srcStride];
1343
        const int src7 = src[7*srcStride];
1344
        const int src8 = src[8*srcStride];
1345
        const int src9 = src[9*srcStride];
1346
        dst[0*dstStride]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
1347
        dst[1*dstStride]= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4];
1348
        dst[2*dstStride]= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4];
1349
        dst[3*dstStride]= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4];
1350
        dst[4*dstStride]= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4];
1351
        dst[5*dstStride]= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4];
1352
        dst[6*dstStride]= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4];
1353
        dst[7*dstStride]= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4];
1354
        src++;
1355
        dst++;
1356
    }
1357
}
1358

    
1359
static void put_mspel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){
1360
    put_pixels8_c(dst, src, stride, 8);
1361
}
1362

    
1363
static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){
1364
    uint8_t half[64];
1365
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1366
    put_pixels8_l2(dst, src, half, stride, stride, 8, 8);
1367
}
1368

    
1369
static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){
1370
    wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
1371
}
1372

    
1373
static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){
1374
    uint8_t half[64];
1375
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1376
    put_pixels8_l2(dst, src+1, half, stride, stride, 8, 8);
1377
}
1378

    
1379
static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){
1380
    wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
1381
}
1382

    
1383
static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){
1384
    uint8_t halfH[88];
1385
    uint8_t halfV[64];
1386
    uint8_t halfHV[64];
1387
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1388
    wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
1389
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1390
    put_pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);
1391
}
1392
static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){
1393
    uint8_t halfH[88];
1394
    uint8_t halfV[64];
1395
    uint8_t halfHV[64];
1396
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1397
    wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
1398
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1399
    put_pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);
1400
}
1401
static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){
1402
    uint8_t halfH[88];
1403
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1404
    wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
1405
}
1406

    
1407

    
1408
static inline int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1409
{
1410
    int s, i;
1411

    
1412
    s = 0;
1413
    for(i=0;i<16;i++) {
1414
        s += abs(pix1[0] - pix2[0]);
1415
        s += abs(pix1[1] - pix2[1]);
1416
        s += abs(pix1[2] - pix2[2]);
1417
        s += abs(pix1[3] - pix2[3]);
1418
        s += abs(pix1[4] - pix2[4]);
1419
        s += abs(pix1[5] - pix2[5]);
1420
        s += abs(pix1[6] - pix2[6]);
1421
        s += abs(pix1[7] - pix2[7]);
1422
        s += abs(pix1[8] - pix2[8]);
1423
        s += abs(pix1[9] - pix2[9]);
1424
        s += abs(pix1[10] - pix2[10]);
1425
        s += abs(pix1[11] - pix2[11]);
1426
        s += abs(pix1[12] - pix2[12]);
1427
        s += abs(pix1[13] - pix2[13]);
1428
        s += abs(pix1[14] - pix2[14]);
1429
        s += abs(pix1[15] - pix2[15]);
1430
        pix1 += line_size;
1431
        pix2 += line_size;
1432
    }
1433
    return s;
1434
}
1435

    
1436
static int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1437
{
1438
    int s, i;
1439

    
1440
    s = 0;
1441
    for(i=0;i<16;i++) {
1442
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
1443
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
1444
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
1445
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
1446
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
1447
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
1448
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
1449
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
1450
        s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
1451
        s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
1452
        s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
1453
        s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
1454
        s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
1455
        s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
1456
        s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
1457
        s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
1458
        pix1 += line_size;
1459
        pix2 += line_size;
1460
    }
1461
    return s;
1462
}
1463

    
1464
static int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1465
{
1466
    int s, i;
1467
    UINT8 *pix3 = pix2 + line_size;
1468

    
1469
    s = 0;
1470
    for(i=0;i<16;i++) {
1471
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
1472
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
1473
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
1474
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
1475
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
1476
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
1477
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
1478
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
1479
        s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
1480
        s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
1481
        s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
1482
        s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
1483
        s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
1484
        s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
1485
        s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
1486
        s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
1487
        pix1 += line_size;
1488
        pix2 += line_size;
1489
        pix3 += line_size;
1490
    }
1491
    return s;
1492
}
1493

    
1494
static int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1495
{
1496
    int s, i;
1497
    UINT8 *pix3 = pix2 + line_size;
1498

    
1499
    s = 0;
1500
    for(i=0;i<16;i++) {
1501
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
1502
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
1503
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
1504
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
1505
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
1506
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
1507
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
1508
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
1509
        s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
1510
        s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
1511
        s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
1512
        s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
1513
        s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
1514
        s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
1515
        s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
1516
        s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
1517
        pix1 += line_size;
1518
        pix2 += line_size;
1519
        pix3 += line_size;
1520
    }
1521
    return s;
1522
}
1523

    
1524
static inline int pix_abs8x8_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1525
{
1526
    int s, i;
1527

    
1528
    s = 0;
1529
    for(i=0;i<8;i++) {
1530
        s += abs(pix1[0] - pix2[0]);
1531
        s += abs(pix1[1] - pix2[1]);
1532
        s += abs(pix1[2] - pix2[2]);
1533
        s += abs(pix1[3] - pix2[3]);
1534
        s += abs(pix1[4] - pix2[4]);
1535
        s += abs(pix1[5] - pix2[5]);
1536
        s += abs(pix1[6] - pix2[6]);
1537
        s += abs(pix1[7] - pix2[7]);
1538
        pix1 += line_size;
1539
        pix2 += line_size;
1540
    }
1541
    return s;
1542
}
1543

    
1544
static int pix_abs8x8_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1545
{
1546
    int s, i;
1547

    
1548
    s = 0;
1549
    for(i=0;i<8;i++) {
1550
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
1551
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
1552
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
1553
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
1554
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
1555
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
1556
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
1557
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
1558
        pix1 += line_size;
1559
        pix2 += line_size;
1560
    }
1561
    return s;
1562
}
1563

    
1564
static int pix_abs8x8_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1565
{
1566
    int s, i;
1567
    UINT8 *pix3 = pix2 + line_size;
1568

    
1569
    s = 0;
1570
    for(i=0;i<8;i++) {
1571
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
1572
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
1573
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
1574
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
1575
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
1576
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
1577
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
1578
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
1579
        pix1 += line_size;
1580
        pix2 += line_size;
1581
        pix3 += line_size;
1582
    }
1583
    return s;
1584
}
1585

    
1586
static int pix_abs8x8_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)
1587
{
1588
    int s, i;
1589
    UINT8 *pix3 = pix2 + line_size;
1590

    
1591
    s = 0;
1592
    for(i=0;i<8;i++) {
1593
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
1594
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
1595
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
1596
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
1597
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
1598
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
1599
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
1600
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
1601
        pix1 += line_size;
1602
        pix2 += line_size;
1603
        pix3 += line_size;
1604
    }
1605
    return s;
1606
}
1607

    
1608
static int sad16x16_c(void *s, uint8_t *a, uint8_t *b, int stride){
1609
    return pix_abs16x16_c(a,b,stride);
1610
}
1611

    
1612
static int sad8x8_c(void *s, uint8_t *a, uint8_t *b, int stride){
1613
    return pix_abs8x8_c(a,b,stride);
1614
}
1615

    
1616
void ff_block_permute(DCTELEM *block, UINT8 *permutation, const UINT8 *scantable, int last)
1617
{
1618
    int i;
1619
    DCTELEM temp[64];
1620
    
1621
    if(last<=0) return;
1622
    //if(permutation[1]==1) return; //FIXME its ok but not clean and might fail for some perms
1623

    
1624
    for(i=0; i<=last; i++){
1625
        const int j= scantable[i];
1626
        temp[j]= block[j];
1627
        block[j]=0;
1628
    }
1629
    
1630
    for(i=0; i<=last; i++){
1631
        const int j= scantable[i];
1632
        const int perm_j= permutation[j];
1633
        block[perm_j]= temp[j];
1634
    }
1635
}
1636

    
1637
static void clear_blocks_c(DCTELEM *blocks)
1638
{
1639
    memset(blocks, 0, sizeof(DCTELEM)*6*64);
1640
}
1641

    
1642
static void add_bytes_c(uint8_t *dst, uint8_t *src, int w){
1643
    int i;
1644
    for(i=0; i+7<w; i+=8){
1645
        dst[i+0] += src[i+0];
1646
        dst[i+1] += src[i+1];
1647
        dst[i+2] += src[i+2];
1648
        dst[i+3] += src[i+3];
1649
        dst[i+4] += src[i+4];
1650
        dst[i+5] += src[i+5];
1651
        dst[i+6] += src[i+6];
1652
        dst[i+7] += src[i+7];
1653
    }
1654
    for(; i<w; i++)
1655
        dst[i+0] += src[i+0];
1656
}
1657

    
1658
static void diff_bytes_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
1659
    int i;
1660
    for(i=0; i+7<w; i+=8){
1661
        dst[i+0] = src1[i+0]-src2[i+0];
1662
        dst[i+1] = src1[i+1]-src2[i+1];
1663
        dst[i+2] = src1[i+2]-src2[i+2];
1664
        dst[i+3] = src1[i+3]-src2[i+3];
1665
        dst[i+4] = src1[i+4]-src2[i+4];
1666
        dst[i+5] = src1[i+5]-src2[i+5];
1667
        dst[i+6] = src1[i+6]-src2[i+6];
1668
        dst[i+7] = src1[i+7]-src2[i+7];
1669
    }
1670
    for(; i<w; i++)
1671
        dst[i+0] = src1[i+0]-src2[i+0];
1672
}
1673

    
1674
#define BUTTERFLY2(o1,o2,i1,i2) \
1675
o1= (i1)+(i2);\
1676
o2= (i1)-(i2);
1677

    
1678
#define BUTTERFLY1(x,y) \
1679
{\
1680
    int a,b;\
1681
    a= x;\
1682
    b= y;\
1683
    x= a+b;\
1684
    y= a-b;\
1685
}
1686

    
1687
#define BUTTERFLYA(x,y) (ABS((x)+(y)) + ABS((x)-(y)))
1688

    
1689
static int hadamard8_diff_c(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride){
1690
    int i;
1691
    int temp[64];
1692
    int sum=0;
1693

    
1694
    for(i=0; i<8; i++){
1695
        //FIXME try pointer walks
1696
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-dst[stride*i+0],src[stride*i+1]-dst[stride*i+1]);
1697
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-dst[stride*i+2],src[stride*i+3]-dst[stride*i+3]);
1698
        BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4]-dst[stride*i+4],src[stride*i+5]-dst[stride*i+5]);
1699
        BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6]-dst[stride*i+6],src[stride*i+7]-dst[stride*i+7]);
1700
        
1701
        BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
1702
        BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
1703
        BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
1704
        BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
1705
        
1706
        BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
1707
        BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
1708
        BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
1709
        BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
1710
    }
1711

    
1712
    for(i=0; i<8; i++){
1713
        BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
1714
        BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
1715
        BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
1716
        BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
1717
        
1718
        BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
1719
        BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
1720
        BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
1721
        BUTTERFLY1(temp[8*5+i], temp[8*7+i]);
1722

    
1723
        sum += 
1724
             BUTTERFLYA(temp[8*0+i], temp[8*4+i])
1725
            +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
1726
            +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
1727
            +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
1728
    }
1729
#if 0
1730
static int maxi=0;
1731
if(sum>maxi){
1732
    maxi=sum;
1733
    printf("MAX:%d\n", maxi);
1734
}
1735
#endif
1736
    return sum;
1737
}
1738

    
1739
static int hadamard8_abs_c(uint8_t *src, int stride, int mean){
1740
    int i;
1741
    int temp[64];
1742
    int sum=0;
1743
//FIXME OOOPS ignore 0 term instead of mean mess
1744
    for(i=0; i<8; i++){
1745
        //FIXME try pointer walks
1746
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-mean,src[stride*i+1]-mean);
1747
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-mean,src[stride*i+3]-mean);
1748
        BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4]-mean,src[stride*i+5]-mean);
1749
        BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6]-mean,src[stride*i+7]-mean);
1750
        
1751
        BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
1752
        BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
1753
        BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
1754
        BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
1755
        
1756
        BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
1757
        BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
1758
        BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
1759
        BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
1760
    }
1761

    
1762
    for(i=0; i<8; i++){
1763
        BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
1764
        BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
1765
        BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
1766
        BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
1767
        
1768
        BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
1769
        BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
1770
        BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
1771
        BUTTERFLY1(temp[8*5+i], temp[8*7+i]);
1772
    
1773
        sum += 
1774
             BUTTERFLYA(temp[8*0+i], temp[8*4+i])
1775
            +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
1776
            +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
1777
            +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
1778
    }
1779
    
1780
    return sum;
1781
}
1782

    
1783
static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
1784
    MpegEncContext * const s= (MpegEncContext *)c;
1785
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
1786
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
1787
    int sum=0, i;
1788

    
1789
    s->dsp.diff_pixels(temp, src1, src2, stride);
1790
    s->fdct(temp);
1791

    
1792
    for(i=0; i<64; i++)
1793
        sum+= ABS(temp[i]);
1794
        
1795
    return sum;
1796
}
1797

    
1798
void simple_idct(DCTELEM *block); //FIXME
1799

    
1800
static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
1801
    MpegEncContext * const s= (MpegEncContext *)c;
1802
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64*2/8];
1803
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
1804
    DCTELEM * const bak = ((DCTELEM*)aligned_temp)+64;
1805
    int sum=0, i;
1806

    
1807
    s->mb_intra=0;
1808
    
1809
    s->dsp.diff_pixels(temp, src1, src2, stride);
1810
    
1811
    memcpy(bak, temp, 64*sizeof(DCTELEM));
1812
    
1813
    s->block_last_index[0/*FIXME*/]= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
1814
    s->dct_unquantize(s, temp, 0, s->qscale);
1815
    simple_idct(temp); //FIXME 
1816
    
1817
    for(i=0; i<64; i++)
1818
        sum+= (temp[i]-bak[i])*(temp[i]-bak[i]);
1819
        
1820
    return sum;
1821
}
1822

    
1823
static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
1824
    MpegEncContext * const s= (MpegEncContext *)c;
1825
    const UINT8 *scantable= s->intra_scantable.permutated;
1826
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
1827
    uint64_t __align8 aligned_bak[stride];
1828
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
1829
    uint8_t * const bak= (uint8_t*)aligned_bak;
1830
    int i, last, run, bits, level, distoration, start_i;
1831
    const int esc_length= s->ac_esc_length;
1832
    uint8_t * length;
1833
    uint8_t * last_length;
1834
    
1835
    for(i=0; i<8; i++){
1836
        ((uint32_t*)(bak + i*stride))[0]= ((uint32_t*)(src2 + i*stride))[0];
1837
        ((uint32_t*)(bak + i*stride))[1]= ((uint32_t*)(src2 + i*stride))[1];
1838
    }
1839

    
1840
    s->dsp.diff_pixels(temp, src1, src2, stride);
1841

    
1842
    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
1843

    
1844
    bits=0;
1845
    
1846
    if (s->mb_intra) {
1847
        start_i = 1; 
1848
        length     = s->intra_ac_vlc_length;
1849
        last_length= s->intra_ac_vlc_last_length;
1850
        bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
1851
    } else {
1852
        start_i = 0;
1853
        length     = s->inter_ac_vlc_length;
1854
        last_length= s->inter_ac_vlc_last_length;
1855
    }
1856
    
1857
    if(last>=start_i){
1858
        run=0;
1859
        for(i=start_i; i<last; i++){
1860
            int j= scantable[i];
1861
            level= temp[j];
1862
        
1863
            if(level){
1864
                level+=64;
1865
                if((level&(~127)) == 0){
1866
                    bits+= length[UNI_AC_ENC_INDEX(run, level)];
1867
                }else
1868
                    bits+= esc_length;
1869
                run=0;
1870
            }else
1871
                run++;
1872
        }
1873
        i= scantable[last];
1874
       
1875
        level= temp[i] + 64;
1876

    
1877
        assert(level - 64);
1878
        
1879
        if((level&(~127)) == 0){
1880
            bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
1881
        }else
1882
            bits+= esc_length;
1883
    
1884
    }
1885

    
1886
    if(last>=0){
1887
        s->dct_unquantize(s, temp, 0, s->qscale);
1888
    }
1889
    
1890
    s->idct_add(bak, stride, temp);
1891
    
1892
    distoration= s->dsp.sse[1](NULL, bak, src1, stride);
1893

    
1894
    return distoration + ((bits*s->qscale*s->qscale*109 + 64)>>7);
1895
}
1896

    
1897
static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
1898
    MpegEncContext * const s= (MpegEncContext *)c;
1899
    const UINT8 *scantable= s->intra_scantable.permutated;
1900
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
1901
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
1902
    int i, last, run, bits, level, start_i;
1903
    const int esc_length= s->ac_esc_length;
1904
    uint8_t * length;
1905
    uint8_t * last_length;
1906
    
1907
    s->dsp.diff_pixels(temp, src1, src2, stride);
1908

    
1909
    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
1910

    
1911
    bits=0;
1912
    
1913
    if (s->mb_intra) {
1914
        start_i = 1; 
1915
        length     = s->intra_ac_vlc_length;
1916
        last_length= s->intra_ac_vlc_last_length;
1917
        bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
1918
    } else {
1919
        start_i = 0;
1920
        length     = s->inter_ac_vlc_length;
1921
        last_length= s->inter_ac_vlc_last_length;
1922
    }
1923
    
1924
    if(last>=start_i){
1925
        run=0;
1926
        for(i=start_i; i<last; i++){
1927
            int j= scantable[i];
1928
            level= temp[j];
1929
        
1930
            if(level){
1931
                level+=64;
1932
                if((level&(~127)) == 0){
1933
                    bits+= length[UNI_AC_ENC_INDEX(run, level)];
1934
                }else
1935
                    bits+= esc_length;
1936
                run=0;
1937
            }else
1938
                run++;
1939
        }
1940
        i= scantable[last];
1941
                
1942
        level= temp[i] + 64;
1943
        
1944
        assert(level - 64);
1945
        
1946
        if((level&(~127)) == 0){
1947
            bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
1948
        }else
1949
            bits+= esc_length;
1950
    }
1951

    
1952
    return bits;
1953
}
1954

    
1955

    
1956
WARPER88_1616(hadamard8_diff_c, hadamard8_diff16_c)
1957
WARPER88_1616(dct_sad8x8_c, dct_sad16x16_c)
1958
WARPER88_1616(quant_psnr8x8_c, quant_psnr16x16_c)
1959
WARPER88_1616(rd8x8_c, rd16x16_c)
1960
WARPER88_1616(bit8x8_c, bit16x16_c)
1961

    
1962
void dsputil_init(DSPContext* c, unsigned mask)
1963
{
1964
    static int init_done = 0;
1965
    int i;
1966

    
1967
    if (!init_done) {
1968
        for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;
1969
        for(i=0;i<MAX_NEG_CROP;i++) {
1970
            cropTbl[i] = 0;
1971
            cropTbl[i + MAX_NEG_CROP + 256] = 255;
1972
        }
1973

    
1974
        for(i=0;i<512;i++) {
1975
            squareTbl[i] = (i - 256) * (i - 256);
1976
        }
1977

    
1978
        for(i=0; i<64; i++) inv_zigzag_direct16[ff_zigzag_direct[i]]= i+1;
1979

    
1980
        init_done = 1;
1981
    }
1982

    
1983
    c->get_pixels = get_pixels_c;
1984
    c->diff_pixels = diff_pixels_c;
1985
    c->put_pixels_clamped = put_pixels_clamped_c;
1986
    c->add_pixels_clamped = add_pixels_clamped_c;
1987
    c->gmc1 = gmc1_c;
1988
    c->gmc = gmc_c;
1989
    c->clear_blocks = clear_blocks_c;
1990
    c->pix_sum = pix_sum_c;
1991
    c->pix_norm1 = pix_norm1_c;
1992
    c->sse[0]= sse16_c;
1993
    c->sse[1]= sse8_c;
1994

    
1995
    /* TODO [0] 16  [1] 8 */
1996
    c->pix_abs16x16     = pix_abs16x16_c;
1997
    c->pix_abs16x16_x2  = pix_abs16x16_x2_c;
1998
    c->pix_abs16x16_y2  = pix_abs16x16_y2_c;
1999
    c->pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
2000
    c->pix_abs8x8     = pix_abs8x8_c;
2001
    c->pix_abs8x8_x2  = pix_abs8x8_x2_c;
2002
    c->pix_abs8x8_y2  = pix_abs8x8_y2_c;
2003
    c->pix_abs8x8_xy2 = pix_abs8x8_xy2_c;
2004

    
2005
#define dspfunc(PFX, IDX, NUM) \
2006
    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c;     \
2007
    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## NUM ## _x2_c;  \
2008
    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## NUM ## _y2_c;  \
2009
    c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## NUM ## _xy2_c
2010

    
2011
    dspfunc(put, 0, 16);
2012
    dspfunc(put_no_rnd, 0, 16);
2013
    dspfunc(put, 1, 8);
2014
    dspfunc(put_no_rnd, 1, 8);
2015

    
2016
    dspfunc(avg, 0, 16);
2017
    dspfunc(avg_no_rnd, 0, 16);
2018
    dspfunc(avg, 1, 8);
2019
    dspfunc(avg_no_rnd, 1, 8);
2020
#undef dspfunc
2021

    
2022
#define dspfunc(PFX, IDX, NUM) \
2023
    c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
2024
    c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
2025
    c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
2026
    c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
2027
    c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
2028
    c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
2029
    c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
2030
    c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
2031
    c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
2032
    c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
2033
    c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
2034
    c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
2035
    c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
2036
    c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
2037
    c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
2038
    c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c
2039

    
2040
    dspfunc(put_qpel, 0, 16);
2041
    dspfunc(put_no_rnd_qpel, 0, 16);
2042

    
2043
    dspfunc(avg_qpel, 0, 16);
2044
    /* dspfunc(avg_no_rnd_qpel, 0, 16); */
2045

    
2046
    dspfunc(put_qpel, 1, 8);
2047
    dspfunc(put_no_rnd_qpel, 1, 8);
2048

    
2049
    dspfunc(avg_qpel, 1, 8);
2050
    /* dspfunc(avg_no_rnd_qpel, 1, 8); */
2051
#undef dspfunc
2052

    
2053
    c->put_mspel_pixels_tab[0]= put_mspel8_mc00_c;
2054
    c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c;
2055
    c->put_mspel_pixels_tab[2]= put_mspel8_mc20_c;
2056
    c->put_mspel_pixels_tab[3]= put_mspel8_mc30_c;
2057
    c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
2058
    c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
2059
    c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
2060
    c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;
2061
    
2062
    c->hadamard8_diff[0]= hadamard8_diff16_c;
2063
    c->hadamard8_diff[1]= hadamard8_diff_c;
2064
    c->hadamard8_abs = hadamard8_abs_c;
2065
    
2066
    c->dct_sad[0]= dct_sad16x16_c;
2067
    c->dct_sad[1]= dct_sad8x8_c;
2068
    
2069
    c->sad[0]= sad16x16_c;
2070
    c->sad[1]= sad8x8_c;
2071
    
2072
    c->quant_psnr[0]= quant_psnr16x16_c;
2073
    c->quant_psnr[1]= quant_psnr8x8_c;
2074

    
2075
    c->rd[0]= rd16x16_c;
2076
    c->rd[1]= rd8x8_c;
2077

    
2078
    c->bit[0]= bit16x16_c;
2079
    c->bit[1]= bit8x8_c;
2080
        
2081
    c->add_bytes= add_bytes_c;
2082
    c->diff_bytes= diff_bytes_c;
2083

    
2084
#ifdef HAVE_MMX
2085
    dsputil_init_mmx(c, mask);
2086
    if (ff_bit_exact)
2087
    {
2088
        /* FIXME - AVCodec context should have flag for bitexact match */
2089
        /* fprintf(stderr, "\n\n\nff_bit_exact %d\n\n\n\n", ff_bit_exact); */
2090
        dsputil_set_bit_exact_mmx(c, mask);
2091
    }
2092
#endif
2093
#ifdef ARCH_ARMV4L
2094
    dsputil_init_armv4l(c, mask);
2095
#endif
2096
#ifdef HAVE_MLIB
2097
    dsputil_init_mlib(c, mask);
2098
#endif
2099
#ifdef ARCH_ALPHA
2100
    dsputil_init_alpha(c, mask);
2101
#endif
2102
#ifdef ARCH_POWERPC
2103
    dsputil_init_ppc(c, mask);
2104
#endif
2105
#ifdef HAVE_MMI
2106
    dsputil_init_mmi(c, mask);
2107
#endif
2108
}
2109

    
2110
/* remove any non bit exact operation (testing purpose) */
2111
void avcodec_set_bit_exact(void)
2112
{
2113
    ff_bit_exact=1;
2114
#ifdef HAVE_MMX
2115
// FIXME - better set_bit_exact
2116
//    dsputil_set_bit_exact_mmx();
2117
#endif
2118
}