aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/dsputil.h
blob: 34d659d532bf96a498a8107bd2d4adcab930723d (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
/*
 * DSP utils
 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * DSP utils.
 * note, many functions in here may use MMX which trashes the FPU state, it is
 * absolutely necessary to call emms_c() between dsp & float/double code
 */

#ifndef AVCODEC_DSPUTIL_H
#define AVCODEC_DSPUTIL_H

#include "libavutil/intreadwrite.h"
#include "avcodec.h"


//#define DEBUG
/* dct code */
typedef short DCTELEM;

void ff_fdct_ifast (DCTELEM *data);
void ff_fdct_ifast248 (DCTELEM *data);
void ff_jpeg_fdct_islow_8(DCTELEM *data);
void ff_jpeg_fdct_islow_10(DCTELEM *data);
void ff_fdct248_islow_8(DCTELEM *data);
void ff_fdct248_islow_10(DCTELEM *data);

void ff_j_rev_dct (DCTELEM *data);
void ff_j_rev_dct4 (DCTELEM *data);
void ff_j_rev_dct2 (DCTELEM *data);
void ff_j_rev_dct1 (DCTELEM *data);
void ff_wmv2_idct_c(DCTELEM *data);

void ff_fdct_mmx(DCTELEM *block);
void ff_fdct_mmx2(DCTELEM *block);
void ff_fdct_sse2(DCTELEM *block);

#define H264_IDCT(depth) \
void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\
void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\
void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);

H264_IDCT( 8)
H264_IDCT( 9)
H264_IDCT(10)

void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);

/* encoding scans */
extern const uint8_t ff_alternate_horizontal_scan[64];
extern const uint8_t ff_alternate_vertical_scan[64];
extern const uint8_t ff_zigzag_direct[64];
extern const uint8_t ff_zigzag248_direct[64];

/* pixel operations */
#define MAX_NEG_CROP 1024

/* temporary */
extern uint32_t ff_squareTbl[512];
extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];

#define PUTAVG_PIXELS(depth)\
void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);

PUTAVG_PIXELS( 8)
PUTAVG_PIXELS( 9)
PUTAVG_PIXELS(10)

#define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
#define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
#define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
#define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c

/* VP3 DSP functions */
void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);

void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);

/* EA functions */
void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);

/* RV40 functions */
void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);

/* 1/2^n downscaling functions from imgconvert.c */
void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);

void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
              int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);

/* minimum alignment rules ;)
If you notice errors in the align stuff, need more alignment for some ASM code
for some CPU or need to use a function with less aligned data then send a mail
to the ffmpeg-devel mailing list, ...

!warning These alignments might not match reality, (missing attribute((align))
stuff somewhere possible).
I (Michael) did not check them, these are just the alignments which I think
could be reached easily ...

!future video codecs might need functions with less strict alignment
*/

/*
void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
void clear_blocks_c(DCTELEM *blocks);
*/

/* add and put pixel (decoding) */
// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);

typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);

#define DEF_OLD_QPEL(name)\
void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);

DEF_OLD_QPEL(qpel16_mc11_old_c)
DEF_OLD_QPEL(qpel16_mc31_old_c)
DEF_OLD_QPEL(qpel16_mc12_old_c)
DEF_OLD_QPEL(qpel16_mc32_old_c)
DEF_OLD_QPEL(qpel16_mc13_old_c)
DEF_OLD_QPEL(qpel16_mc33_old_c)
DEF_OLD_QPEL(qpel8_mc11_old_c)
DEF_OLD_QPEL(qpel8_mc31_old_c)
DEF_OLD_QPEL(qpel8_mc12_old_c)
DEF_OLD_QPEL(qpel8_mc32_old_c)
DEF_OLD_QPEL(qpel8_mc13_old_c)
DEF_OLD_QPEL(qpel8_mc33_old_c)

#define CALL_2X_PIXELS(a, b, n)\
static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    b(block  , pixels  , line_size, h);\
    b(block+n, pixels+n, line_size, h);\
}

/* motion estimation */
// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
// although currently h<4 is not used as functions with width <8 are neither used nor implemented
typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;

/**
 * Scantable.
 */
typedef struct ScanTable{
    const uint8_t *scantable;
    uint8_t permutated[64];
    uint8_t raster_end[64];
} ScanTable;

void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
void ff_init_scantable_permutation(uint8_t *idct_permutation,
                                   int idct_permutation_type);

#define EMULATED_EDGE(depth) \
void ff_emulated_edge_mc_ ## depth (uint8_t *buf, const uint8_t *src, int linesize,\
                         int block_w, int block_h,\
                         int src_x, int src_y, int w, int h);

EMULATED_EDGE(8)
EMULATED_EDGE(9)
EMULATED_EDGE(10)

void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);

/**
 * DSPContext.
 */
typedef struct DSPContext {
    /**
     * Size of DCT coefficients.
     */
    int dct_bits;

    /* pixel ops : interface with DCT */
    void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
    void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
    void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
    void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
    int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
    /**
     * Motion estimation with emulated edge values.
     * @param buf pointer to destination buffer (unaligned)
     * @param src pointer to pixel source (unaligned)
     * @param linesize width (in pixels) for src/buf
     * @param block_w number of pixels (per row) to copy to buf
     * @param block_h nummber of pixel rows to copy to buf
     * @param src_x offset of src to start of row - this may be negative
     * @param src_y offset of src to top of image - this may be negative
     * @param w width of src in pixels
     * @param h height of src in pixels
     */
    void (*emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize,
                             int block_w, int block_h,
                             int src_x, int src_y, int w, int h);
    /**
     * translational global motion compensation.
     */
    void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
    /**
     * global motion compensation.
     */
    void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
                    int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
    void (*clear_block)(DCTELEM *block/*align 16*/);
    void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
    int (*pix_sum)(uint8_t * pix, int line_size);
    int (*pix_norm1)(uint8_t * pix, int line_size);
// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4

    me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
    me_cmp_func sse[6];
    me_cmp_func hadamard8_diff[6];
    me_cmp_func dct_sad[6];
    me_cmp_func quant_psnr[6];
    me_cmp_func bit[6];
    me_cmp_func rd[6];
    me_cmp_func vsad[6];
    me_cmp_func vsse[6];
    me_cmp_func nsse[6];
    me_cmp_func w53[6];
    me_cmp_func w97[6];
    me_cmp_func dct_max[6];
    me_cmp_func dct264_sad[6];

    me_cmp_func me_pre_cmp[6];
    me_cmp_func me_cmp[6];
    me_cmp_func me_sub_cmp[6];
    me_cmp_func mb_cmp[6];
    me_cmp_func ildct_cmp[6]; //only width 16 used
    me_cmp_func frame_skip_cmp[6]; //only width 8 used

    int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
                             int size);

    /**
     * Halfpel motion compensation with rounding (a+b+1)>>1.
     * this is an array[4][4] of motion compensation functions for 4
     * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
     * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
     * @param block destination where the result is stored
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    op_pixels_func put_pixels_tab[4][4];

    /**
     * Halfpel motion compensation with rounding (a+b+1)>>1.
     * This is an array[4][4] of motion compensation functions for 4
     * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
     * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
     * @param block destination into which the result is averaged (a+b+1)>>1
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    op_pixels_func avg_pixels_tab[4][4];

    /**
     * Halfpel motion compensation with no rounding (a+b)>>1.
     * this is an array[2][4] of motion compensation functions for 2
     * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
     * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
     * @param block destination where the result is stored
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    op_pixels_func put_no_rnd_pixels_tab[4][4];

    /**
     * Halfpel motion compensation with no rounding (a+b)>>1.
     * this is an array[2][4] of motion compensation functions for 2
     * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
     * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
     * @param block destination into which the result is averaged (a+b)>>1
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    op_pixels_func avg_no_rnd_pixels_tab[4][4];

    void (*put_no_rnd_pixels_l2[2])(uint8_t *block/*align width (8 or 16)*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);

    /**
     * Thirdpel motion compensation with rounding (a+b+1)>>1.
     * this is an array[12] of motion compensation functions for the 9 thirdpe
     * positions<br>
     * *pixels_tab[ xthirdpel + 4*ythirdpel ]
     * @param block destination where the result is stored
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
    tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?

    qpel_mc_func put_qpel_pixels_tab[2][16];
    qpel_mc_func avg_qpel_pixels_tab[2][16];
    qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
    qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
    qpel_mc_func put_mspel_pixels_tab[8];

    /**
     * h264 Chroma MC
     */
    h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
    h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];

    qpel_mc_func put_h264_qpel_pixels_tab[4][16];
    qpel_mc_func avg_h264_qpel_pixels_tab[4][16];

    qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
    qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];

    me_cmp_func pix_abs[2][4];

    /* huffyuv specific */
    void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
    void (*diff_bytes)(uint8_t *dst/*align 16*/, const uint8_t *src1/*align 16*/, const uint8_t *src2/*align 1*/,int w);
    /**
     * subtract huffyuv's variant of median prediction
     * note, this might read from src1[-1], src2[-1]
     */
    void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
    void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
    int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
    void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
    /* this might write to dst[w] */
    void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
    void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);

    void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
    void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);

    void (*h261_loop_filter)(uint8_t *src, int stride);

    void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
    void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);

    void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
    void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
    void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);

    /* assume len is a multiple of 4, and arrays are 16-byte aligned */
    void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
    void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
    /* assume len is a multiple of 8, and arrays are 16-byte aligned */
    void (*vector_fmul)(float *dst, const float *src0, const float *src1, int len);
    void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
    /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
    void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
    /* assume len is a multiple of 4, and arrays are 16-byte aligned */
    void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len);
    /* assume len is a multiple of 8, and arrays are 16-byte aligned */
    void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
    /**
     * Multiply a vector of floats by a scalar float.  Source and
     * destination vectors must overlap exactly or not at all.
     * @param dst result vector, 16-byte aligned
     * @param src input vector, 16-byte aligned
     * @param mul scalar value
     * @param len length of vector, multiple of 4
     */
    void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
                               int len);
    /**
     * Multiply a vector of floats by a scalar float and add to
     * destination vector.  Source and destination vectors must
     * overlap exactly or not at all.
     * @param dst result vector, 16-byte aligned
     * @param src input vector, 16-byte aligned
     * @param mul scalar value
     * @param len length of vector, multiple of 4
     */
    void (*vector_fmac_scalar)(float *dst, const float *src, float mul,
                               int len);
    /**
     * Calculate the scalar product of two vectors of floats.
     * @param v1  first vector, 16-byte aligned
     * @param v2  second vector, 16-byte aligned
     * @param len length of vectors, multiple of 4
     */
    float (*scalarproduct_float)(const float *v1, const float *v2, int len);
    /**
     * Calculate the sum and difference of two vectors of floats.
     * @param v1  first input vector, sum output, 16-byte aligned
     * @param v2  second input vector, difference output, 16-byte aligned
     * @param len length of vectors, multiple of 4
     */
    void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);

    /**
     * Calculate the sum and difference of two vectors of floats and interleave
     * results into a separate output vector of floats, with each sum
     * positioned before the corresponding difference.
     *
     * @param dst  output vector
     *             constraints: 16-byte aligned
     * @param src0 first input vector
     *             constraints: 32-byte aligned
     * @param src1 second input vector
     *             constraints: 32-byte aligned
     * @param len  number of elements in the input
     *             constraints: multiple of 8
     */
    void (*butterflies_float_interleave)(float *dst, const float *src0,
                                         const float *src1, int len);

    /* (I)DCT */
    void (*fdct)(DCTELEM *block/* align 16*/);
    void (*fdct248)(DCTELEM *block/* align 16*/);

    /* IDCT really*/
    void (*idct)(DCTELEM *block/* align 16*/);

    /**
     * block -> idct -> clip to unsigned 8 bit -> dest.
     * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);

    /**
     * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);

    /**
     * idct input permutation.
     * several optimized IDCTs need a permutated input (relative to the normal order of the reference
     * IDCT)
     * this permutation must be performed before the idct_put/add, note, normally this can be merged
     * with the zigzag/alternate scan<br>
     * an example to avoid confusion:
     * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
     * - (x -> reference dct -> reference idct -> x)
     * - (x -> reference dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
     * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
     */
    uint8_t idct_permutation[64];
    int idct_permutation_type;
#define FF_NO_IDCT_PERM 1
#define FF_LIBMPEG2_IDCT_PERM 2
#define FF_SIMPLE_IDCT_PERM 3
#define FF_TRANSPOSE_IDCT_PERM 4
#define FF_PARTTRANS_IDCT_PERM 5
#define FF_SSE2_IDCT_PERM 6

    int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
    void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
#define BASIS_SHIFT 16
#define RECON_SHIFT 6

    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
#define EDGE_WIDTH 16
#define EDGE_TOP    1
#define EDGE_BOTTOM 2

    void (*prefetch)(void *mem, int stride, int h);

    void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);

    /* mlp/truehd functions */
    void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
                               int firorder, int iirorder,
                               unsigned int filter_shift, int32_t mask, int blocksize,
                               int32_t *sample_buffer);

    /* intrax8 functions */
    void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
    void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
           int * range, int * sum,  int edges);

    /**
     * Calculate scalar product of two vectors.
     * @param len length of vectors, should be multiple of 16
     */
    int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
    /* ape functions */
    /**
     * Calculate scalar product of v1 and v2,
     * and v1[i] += v3[i] * mul
     * @param len length of vectors, should be multiple of 16
     */
    int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);

    /**
     * Apply symmetric window in 16-bit fixed-point.
     * @param output destination array
     *               constraints: 16-byte aligned
     * @param input  source array
     *               constraints: 16-byte aligned
     * @param window window array
     *               constraints: 16-byte aligned, at least len/2 elements
     * @param len    full window length
     *               constraints: multiple of ? greater than zero
     */
    void (*apply_window_int16)(int16_t *output, const int16_t *input,
                               const int16_t *window, unsigned int len);

    /**
     * Clip each element in an array of int32_t to a given minimum and maximum value.
     * @param dst  destination array
     *             constraints: 16-byte aligned
     * @param src  source array
     *             constraints: 16-byte aligned
     * @param min  minimum value
     *             constraints: must in the the range [-(1<<24), 1<<24]
     * @param max  maximum value
     *             constraints: must in the the range [-(1<<24), 1<<24]
     * @param len  number of elements in the array
     *             constraints: multiple of 32 greater than zero
     */
    void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
                              int32_t max, unsigned int len);

    op_fill_func fill_block_tab[2];
} DSPContext;

void ff_dsputil_static_init(void);
void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
attribute_deprecated void dsputil_init(DSPContext* c, AVCodecContext *avctx);

int ff_check_alignment(void);

/**
 * permute block according to permuatation.
 * @param last last non zero element in scantable order
 */
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);

void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);

#define         BYTE_VEC32(c)   ((c)*0x01010101UL)
#define         BYTE_VEC64(c)   ((c)*0x0001000100010001UL)

static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
{
    return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
}

static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
{
    return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
}

static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
{
    return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
}

static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
{
    return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
}

static inline int get_penalty_factor(int lambda, int lambda2, int type){
    switch(type&0xFF){
    default:
    case FF_CMP_SAD:
        return lambda>>FF_LAMBDA_SHIFT;
    case FF_CMP_DCT:
        return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
    case FF_CMP_W53:
        return (4*lambda)>>(FF_LAMBDA_SHIFT);
    case FF_CMP_W97:
        return (2*lambda)>>(FF_LAMBDA_SHIFT);
    case FF_CMP_SATD:
    case FF_CMP_DCT264:
        return (2*lambda)>>FF_LAMBDA_SHIFT;
    case FF_CMP_RD:
    case FF_CMP_PSNR:
    case FF_CMP_SSE:
    case FF_CMP_NSSE:
        return lambda2>>FF_LAMBDA_SHIFT;
    case FF_CMP_BIT:
        return 1;
    }
}

void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);

void ff_dsputil_init_dwt(DSPContext *c);
void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);

#if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMI || HAVE_MMX
#   define STRIDE_ALIGN 16
#else
#   define STRIDE_ALIGN 8
#endif

#define LOCAL_ALIGNED_A(a, t, v, s, o, ...)             \
    uint8_t la_##v[sizeof(t s o) + (a)];                \
    t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)

#define LOCAL_ALIGNED_D(a, t, v, s, o, ...) DECLARE_ALIGNED(a, t, v) s o

#define LOCAL_ALIGNED(a, t, v, ...) LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,)

#if HAVE_LOCAL_ALIGNED_8
#   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,)
#else
#   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
#endif

#if HAVE_LOCAL_ALIGNED_16
#   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,)
#else
#   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
#endif

#define WRAPPER8_16(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
    return name8(s, dst           , src           , stride, h)\
          +name8(s, dst+8         , src+8         , stride, h);\
}

#define WRAPPER8_16_SQ(name8, name16)\
static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
    int score=0;\
    score +=name8(s, dst           , src           , stride, 8);\
    score +=name8(s, dst+8         , src+8         , stride, 8);\
    if(h==16){\
        dst += 8*stride;\
        src += 8*stride;\
        score +=name8(s, dst           , src           , stride, 8);\
        score +=name8(s, dst+8         , src+8         , stride, 8);\
    }\
    return score;\
}


static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN16(dst   , AV_RN16(src   ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        dst+=dstStride;
        src+=srcStride;
    }
}

static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        dst[8]= src[8];
        dst+=dstStride;
        src+=srcStride;
    }
}

static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        AV_WN32(dst+8 , AV_RN32(src+8 ));
        AV_WN32(dst+12, AV_RN32(src+12));
        dst+=dstStride;
        src+=srcStride;
    }
}

static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
{
    int i;
    for(i=0; i<h; i++)
    {
        AV_WN32(dst   , AV_RN32(src   ));
        AV_WN32(dst+4 , AV_RN32(src+4 ));
        AV_WN32(dst+8 , AV_RN32(src+8 ));
        AV_WN32(dst+12, AV_RN32(src+12));
        dst[16]= src[16];
        dst+=dstStride;
        src+=srcStride;
    }
}

#endif /* AVCODEC_DSPUTIL_H */