dsputil.h
Go to the documentation of this file.
1 /*
2  * DSP utils
3  * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
4  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
5  *
6  * This file is part of Libav.
7  *
8  * Libav is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * Libav is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with Libav; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  */
22 
30 #ifndef AVCODEC_DSPUTIL_H
31 #define AVCODEC_DSPUTIL_H
32 
33 #include "libavutil/intreadwrite.h"
34 #include "avcodec.h"
35 
36 
37 //#define DEBUG
38 /* dct code */
39 typedef short DCTELEM;
40 
41 void ff_fdct_ifast (DCTELEM *data);
47 
48 void ff_j_rev_dct (DCTELEM *data);
50 
54 
55 #define H264_IDCT(depth) \
56 void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
57 void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
58 void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
59 void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
60 void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
61 void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
62 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
63 void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
64 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
65 void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\
66 void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\
67 void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);
68 
69 H264_IDCT( 8)
70 H264_IDCT( 9)
71 H264_IDCT(10)
72 
73 void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
74 void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
75 
76 /* encoding scans */
77 extern const uint8_t ff_alternate_horizontal_scan[64];
78 extern const uint8_t ff_alternate_vertical_scan[64];
79 extern const uint8_t ff_zigzag_direct[64];
80 extern const uint8_t ff_zigzag248_direct[64];
81 
82 /* pixel operations */
83 #define MAX_NEG_CROP 1024
84 
85 /* temporary */
86 extern uint32_t ff_squareTbl[512];
87 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
88 
89 #define PUTAVG_PIXELS(depth)\
90 void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
91 void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
92 void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
93 void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);
94 
95 PUTAVG_PIXELS( 8)
96 PUTAVG_PIXELS( 9)
97 PUTAVG_PIXELS(10)
98 
99 #define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
100 #define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
101 #define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
102 #define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c
103 
104 /* RV40 functions */
105 void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
106 void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
107 void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
108 void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
109 
110 /* 1/2^n downscaling functions from imgconvert.c */
111 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
112 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
113 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
114 
115 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
116  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
117 
118 /* minimum alignment rules ;)
119 If you notice errors in the align stuff, need more alignment for some ASM code
120 for some CPU or need to use a function with less aligned data then send a mail
121 to the libav-devel mailing list, ...
122 
123 !warning These alignments might not match reality, (missing attribute((align))
124 stuff somewhere possible).
125 I (Michael) did not check them, these are just the alignments which I think
126 could be reached easily ...
127 
128 !future video codecs might need functions with less strict alignment
129 */
130 
131 /*
132 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
133 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
134 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
135 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
136 void clear_blocks_c(DCTELEM *blocks);
137 */
138 
139 /* add and put pixel (decoding) */
140 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
141 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
142 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
143 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);
144 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
145 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
146 
147 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
148 
149 #define DEF_OLD_QPEL(name)\
150 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
151 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
152 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
153 
154 DEF_OLD_QPEL(qpel16_mc11_old_c)
155 DEF_OLD_QPEL(qpel16_mc31_old_c)
156 DEF_OLD_QPEL(qpel16_mc12_old_c)
157 DEF_OLD_QPEL(qpel16_mc32_old_c)
158 DEF_OLD_QPEL(qpel16_mc13_old_c)
159 DEF_OLD_QPEL(qpel16_mc33_old_c)
160 DEF_OLD_QPEL(qpel8_mc11_old_c)
161 DEF_OLD_QPEL(qpel8_mc31_old_c)
162 DEF_OLD_QPEL(qpel8_mc12_old_c)
163 DEF_OLD_QPEL(qpel8_mc32_old_c)
164 DEF_OLD_QPEL(qpel8_mc13_old_c)
165 DEF_OLD_QPEL(qpel8_mc33_old_c)
166 
167 #define CALL_2X_PIXELS(a, b, n)\
168 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
169  b(block , pixels , line_size, h);\
170  b(block+n, pixels+n, line_size, h);\
171 }
172 
173 /* motion estimation */
174 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
175 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
176 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))*/;
177 
181 typedef struct ScanTable{
185 } ScanTable;
186 
187 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
188 void ff_init_scantable_permutation(uint8_t *idct_permutation,
189  int idct_permutation_type);
190 
194 typedef struct DSPContext {
198  int dct_bits;
199 
200  /* pixel ops : interface with DCT */
201  void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
202  void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
203  void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
204  void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
205  void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
206  void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
207  void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
208  int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
212  void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
216  void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
217  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
218  void (*clear_block)(DCTELEM *block/*align 16*/);
219  void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
220  int (*pix_sum)(uint8_t * pix, int line_size);
221  int (*pix_norm1)(uint8_t * pix, int line_size);
222 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
223 
224  me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
238 
243  me_cmp_func ildct_cmp[6]; //only width 16 used
244  me_cmp_func frame_skip_cmp[6]; //only width 8 used
245 
246  int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
247  int size);
248 
260 
272 
284 
296 
297  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);
298 
309  tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
310  tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
311 
317 
323 
326 
329 
331 
332  /* huffyuv specific */
333  void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
334  void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
339  void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
340  void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
341  int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
342  void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
343  void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
344  void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
345 
346  void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
347  void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
348 
350 
351  /* assume len is a multiple of 4, and arrays are 16-byte aligned */
352  void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
353  /* assume len is a multiple of 16, and arrays are 32-byte aligned */
354  void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
355  /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
356  void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
357  /* assume len is a multiple of 4, and arrays are 16-byte aligned */
358  void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len);
359  /* assume len is a multiple of 8, and arrays are 16-byte aligned */
360  void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
367  float (*scalarproduct_float)(const float *v1, const float *v2, int len);
374  void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
375 
390  void (*butterflies_float_interleave)(float *dst, const float *src0,
391  const float *src1, int len);
392 
393  /* (I)DCT */
394  void (*fdct)(DCTELEM *block/* align 16*/);
395  void (*fdct248)(DCTELEM *block/* align 16*/);
396 
397  /* IDCT really*/
398  void (*idct)(DCTELEM *block/* align 16*/);
399 
405  void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
406 
411  void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
412 
427 #define FF_NO_IDCT_PERM 1
428 #define FF_LIBMPEG2_IDCT_PERM 2
429 #define FF_SIMPLE_IDCT_PERM 3
430 #define FF_TRANSPOSE_IDCT_PERM 4
431 #define FF_PARTTRANS_IDCT_PERM 5
432 #define FF_SSE2_IDCT_PERM 6
433 
434  int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
435  void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
436 #define BASIS_SHIFT 16
437 #define RECON_SHIFT 6
438 
439  void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
440 #define EDGE_WIDTH 16
441 #define EDGE_TOP 1
442 #define EDGE_BOTTOM 2
443 
444  void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
445 
450  int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
451  /* ape functions */
457  int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
458 
470  void (*apply_window_int16)(int16_t *output, const int16_t *input,
471  const int16_t *window, unsigned int len);
472 
487  int32_t max, unsigned int len);
488 
490 } DSPContext;
491 
492 void ff_dsputil_static_init(void);
493 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
494 
495 int ff_check_alignment(void);
496 
506 float ff_scalarproduct_float_c(const float *v1, const float *v2, int len);
507 
512 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
513 
514 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
515 
516 #define BYTE_VEC32(c) ((c)*0x01010101UL)
517 #define BYTE_VEC64(c) ((c)*0x0001000100010001UL)
518 
519 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
520 {
521  return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
522 }
523 
524 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
525 {
526  return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
527 }
528 
529 static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
530 {
531  return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
532 }
533 
534 static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
535 {
536  return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
537 }
538 
539 static inline int get_penalty_factor(int lambda, int lambda2, int type){
540  switch(type&0xFF){
541  default:
542  case FF_CMP_SAD:
543  return lambda>>FF_LAMBDA_SHIFT;
544  case FF_CMP_DCT:
545  return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
546  case FF_CMP_W53:
547  return (4*lambda)>>(FF_LAMBDA_SHIFT);
548  case FF_CMP_W97:
549  return (2*lambda)>>(FF_LAMBDA_SHIFT);
550  case FF_CMP_SATD:
551  case FF_CMP_DCT264:
552  return (2*lambda)>>FF_LAMBDA_SHIFT;
553  case FF_CMP_RD:
554  case FF_CMP_PSNR:
555  case FF_CMP_SSE:
556  case FF_CMP_NSSE:
557  return lambda2>>FF_LAMBDA_SHIFT;
558  case FF_CMP_BIT:
559  return 1;
560  }
561 }
562 
570 
572 
573 #if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMX
574 # define STRIDE_ALIGN 16
575 #else
576 # define STRIDE_ALIGN 8
577 #endif
578 
579 // Some broken preprocessors need a second expansion
580 // to be forced to tokenize __VA_ARGS__
581 #define E(x) x
582 
583 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...) \
584  uint8_t la_##v[sizeof(t s o) + (a)]; \
585  t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
586 
587 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) \
588  DECLARE_ALIGNED(a, t, la_##v) s o; \
589  t (*v) o = la_##v
590 
591 #define LOCAL_ALIGNED(a, t, v, ...) E(LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,))
592 
593 #if HAVE_LOCAL_ALIGNED_8
594 # define LOCAL_ALIGNED_8(t, v, ...) E(LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,))
595 #else
596 # define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
597 #endif
598 
599 #if HAVE_LOCAL_ALIGNED_16
600 # define LOCAL_ALIGNED_16(t, v, ...) E(LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,))
601 #else
602 # define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
603 #endif
604 
605 #define WRAPPER8_16_SQ(name8, name16)\
606 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
607  int score=0;\
608  score +=name8(s, dst , src , stride, 8);\
609  score +=name8(s, dst+8 , src+8 , stride, 8);\
610  if(h==16){\
611  dst += 8*stride;\
612  src += 8*stride;\
613  score +=name8(s, dst , src , stride, 8);\
614  score +=name8(s, dst+8 , src+8 , stride, 8);\
615  }\
616  return score;\
617 }
618 
619 
620 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
621 {
622  int i;
623  for(i=0; i<h; i++)
624  {
625  AV_COPY16U(dst, src);
626  dst+=dstStride;
627  src+=srcStride;
628  }
629 }
630 
631 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
632 {
633  int i;
634  for(i=0; i<h; i++)
635  {
636  AV_COPY32U(dst, src);
637  dst+=dstStride;
638  src+=srcStride;
639  }
640 }
641 
642 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
643 {
644  int i;
645  for(i=0; i<h; i++)
646  {
647  AV_COPY64U(dst, src);
648  dst+=dstStride;
649  src+=srcStride;
650  }
651 }
652 
653 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
654 {
655  int i;
656  for(i=0; i<h; i++)
657  {
658  AV_COPY64U(dst, src);
659  dst[8]= src[8];
660  dst+=dstStride;
661  src+=srcStride;
662  }
663 }
664 
665 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
666 {
667  int i;
668  for(i=0; i<h; i++)
669  {
670  AV_COPY128U(dst, src);
671  dst+=dstStride;
672  src+=srcStride;
673  }
674 }
675 
676 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
677 {
678  int i;
679  for(i=0; i<h; i++)
680  {
681  AV_COPY128U(dst, src);
682  dst[16]= src[16];
683  dst+=dstStride;
684  src+=srcStride;
685  }
686 }
687 
688 #endif /* AVCODEC_DSPUTIL_H */
void(* vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len)
Definition: dsputil.h:358
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last)
permute block according to permuatation.
Definition: dsputil.c:1716
void ff_dsputil_static_init(void)
Definition: dsputil.c:2619
me_cmp_func vsad[6]
Definition: dsputil.h:231
op_pixels_func put_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
Definition: dsputil.h:259
me_cmp_func frame_skip_cmp[6]
Definition: dsputil.h:244
int size
void(* add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha)
Definition: dsputil.h:342
void(* shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
Definition: dsputil.h:444
const uint8_t ff_alternate_vertical_scan[64]
Definition: dsputil.c:97
void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil_ppc.c:140
void(* idct_add)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
Definition: dsputil.h:411
void(* vorbis_inverse_coupling)(float *mag, float *ang, int blocksize)
Definition: dsputil.h:352
void ff_dsputil_init_arm(DSPContext *c, AVCodecContext *avctx)
void ff_fdct248_islow_10(DCTELEM *data)
int(* try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale)
Definition: dsputil.h:434
void(* draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
Definition: dsputil.h:439
int(* me_cmp_func)(void *s, uint8_t *blk1, uint8_t *blk2, int line_size, int h)
Definition: dsputil.h:176
Scantable.
Definition: dsputil.h:181
me_cmp_func dct_max[6]
Definition: dsputil.h:236
#define MAX_NEG_CROP
Definition: dsputil.h:83
me_cmp_func sse[6]
Definition: dsputil.h:225
const uint8_t ff_zigzag248_direct[64]
Definition: dsputil.c:72
uint8_t raster_end[64]
Definition: dsputil.h:184
#define wrap(func)
Definition: w64xmmtest.h:70
void(* apply_window_int16)(int16_t *output, const int16_t *input, const int16_t *window, unsigned int len)
Apply symmetric window in 16-bit fixed-point.
Definition: dsputil.h:470
uint8_t permutated[64]
Definition: dsputil.h:183
void ff_j_rev_dct(DCTELEM *data)
int stride
Definition: mace.c:144
void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc)
Definition: svq3.c:183
int(* pix_sum)(uint8_t *pix, int line_size)
Definition: dsputil.h:220
void ff_dsputil_init(DSPContext *p, AVCodecContext *avctx)
Definition: dsputil.c:2656
qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16]
Definition: dsputil.h:315
static uint64_t rnd_avg64(uint64_t a, uint64_t b)
Definition: dsputil.h:529
uint8_t
void(* qpel_mc_func)(uint8_t *dst, uint8_t *src, int stride)
Definition: dsputil.h:144
h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]
Definition: dsputil.h:322
void ff_fdct_mmx(DCTELEM *block)
void(* get_pixels)(DCTELEM *block, const uint8_t *pixels, int line_size)
Definition: dsputil.h:201
const uint8_t ff_alternate_horizontal_scan[64]
Definition: dsputil.c:86
#define b
Definition: input.c:52
uint32_t ff_squareTbl[512]
Definition: dsputil.c:42
#define DEF_OLD_QPEL(name)
Definition: dsputil.h:149
me_cmp_func me_pre_cmp[6]
Definition: dsputil.h:239
static void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:631
me_cmp_func me_cmp[6]
Definition: dsputil.h:240
void(* add_bytes)(uint8_t *dst, uint8_t *src, int w)
Definition: dsputil.h:333
op_pixels_func avg_pixels_tab[4][4]
Halfpel motion compensation with rounding (a+b+1)>>1.
Definition: dsputil.h:271
const char data[16]
Definition: mxf.c:66
me_cmp_func dct_sad[6]
Definition: dsputil.h:227
void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil_mmx.c:2515
uint8_t idct_permutation[64]
idct input permutation.
Definition: dsputil.h:425
me_cmp_func ildct_cmp[6]
Definition: dsputil.h:243
const uint8_t * scantable
Definition: dsputil.h:182
#define AV_COPY128U(d, s)
Definition: intreadwrite.h:485
static uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
Definition: dsputil.h:534
void(* h263_h_loop_filter)(uint8_t *src, int stride, int qscale)
Definition: dsputil.h:347
qpel_mc_func put_qpel_pixels_tab[2][16]
Definition: dsputil.h:312
#define PUTAVG_PIXELS(depth)
Definition: dsputil.h:89
me_cmp_func nsse[6]
Definition: dsputil.h:233
qpel_mc_func put_mspel_pixels_tab[8]
Definition: dsputil.h:316
#define r
Definition: input.c:51
void(* vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len)
Definition: dsputil.h:356
void(* add_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:205
void ff_dsputil_init_alpha(DSPContext *c, AVCodecContext *avctx)
me_cmp_func w53[6]
Definition: dsputil.h:234
int32_t(* scalarproduct_int16)(const int16_t *v1, const int16_t *v2, int len)
Calculate scalar product of two vectors.
Definition: dsputil.h:450
void(* put_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:203
void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
Definition: imgconvert.c:192
void(* bswap16_buf)(uint16_t *dst, const uint16_t *src, int len)
Definition: dsputil.h:344
#define s2
Definition: regdef.h:39
void(* sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top)
subtract huffyuv's variant of median prediction note, this might read from src1[-1], src2[-1]
Definition: dsputil.h:339
static uint32_t rnd_avg32(uint32_t a, uint32_t b)
Definition: dsputil.h:519
tpel_mc_func avg_tpel_pixels_tab[11]
Definition: dsputil.h:310
void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
int ff_check_alignment(void)
Definition: dsputil.c:2636
#define AV_COPY16U(d, s)
Definition: intreadwrite.h:473
void(* bswap_buf)(uint32_t *dst, const uint32_t *src, int w)
Definition: dsputil.h:343
qpel_mc_func avg_qpel_pixels_tab[2][16]
Definition: dsputil.h:313
void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp)
Definition: svq3.c:148
void(* idct_put)(uint8_t *dest, int line_size, DCTELEM *block)
block -> idct -> clip to unsigned 8 bit -> dest.
Definition: dsputil.h:405
void ff_fdct_sse2(DCTELEM *block)
int(* add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left)
Definition: dsputil.h:341
me_cmp_func mb_cmp[6]
Definition: dsputil.h:242
void(* h264_chroma_mc_func)(uint8_t *dst, uint8_t *src, int srcStride, int h, int x, int y)
Definition: dsputil.h:145
void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type)
Definition: dsputil.c:1741
#define AV_COPY64U(d, s)
Definition: intreadwrite.h:481
op_pixels_func put_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
Definition: dsputil.h:283
void ff_dsputil_init_bfin(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil_bfin.c:198
void(* clear_block)(DCTELEM *block)
Definition: dsputil.h:218
void(* idct)(DCTELEM *block)
Definition: dsputil.h:398
int dct_bits
Size of DCT coefficients.
Definition: dsputil.h:198
static void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:642
h264_chroma_mc_func put_h264_chroma_pixels_tab[3]
h264 Chroma MC
Definition: dsputil.h:321
me_cmp_func vsse[6]
Definition: dsputil.h:232
void ff_dsputil_init_dwt(DSPContext *c)
Definition: dwt.c:845
void(* add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size)
Definition: dsputil.h:206
float ff_scalarproduct_float_c(const float *v1, const float *v2, int len)
Return the scalar product of two vectors.
Definition: dsputil.c:2418
void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
void(* add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top)
Definition: dsputil.h:340
static DCTELEM block[64]
Definition: dct-test.c:169
void(* fdct248)(DCTELEM *block)
Definition: dsputil.h:395
void ff_fdct_ifast(DCTELEM *data)
Definition: jfdctfst.c:208
void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
Definition: imgconvert.c:256
qpel_mc_func avg_2tap_qpel_pixels_tab[4][16]
Definition: dsputil.h:328
int32_t
void ff_dsputil_init_vis(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil_vis.c:3953
#define H264_IDCT(depth)
Definition: dsputil.h:55
#define AV_COPY32U(d, s)
Definition: intreadwrite.h:477
static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby, const int size, const int h, int ref_index, int src_index, me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags)
compares a block (either a full macroblock or a partition thereof) against a proposed motion-compensa...
Definition: motion_est.c:252
me_cmp_func bit[6]
Definition: dsputil.h:229
void(* op_pixels_func)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
Definition: dsputil.h:142
void ff_jpeg_fdct_islow_10(DCTELEM *data)
static int width
Definition: utils.c:156
op_pixels_func avg_no_rnd_pixels_tab[4][4]
Halfpel motion compensation with no rounding (a+b)>>1.
Definition: dsputil.h:295
#define BYTE_VEC32(c)
Definition: dsputil.h:516
external API header
int idct_permutation_type
Definition: dsputil.h:426
void(* vector_clipf)(float *dst, const float *src, float min, float max, int len)
Definition: dsputil.h:360
static int get_penalty_factor(int lambda, int lambda2, int type)
Definition: dsputil.h:539
main external API structure.
Definition: avcodec.h:1339
static void(WINAPI *cond_broadcast)(pthread_cond_t *cond)
void(* butterflies_float)(float *restrict v1, float *restrict v2, int len)
Calculate the sum and difference of two vectors of floats.
Definition: dsputil.h:374
me_cmp_func dct264_sad[6]
Definition: dsputil.h:237
void(* diff_pixels)(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride)
Definition: dsputil.h:202
op_fill_func fill_block_tab[2]
Definition: dsputil.h:489
void(* gmc1)(uint8_t *dst, uint8_t *src, int srcStride, int h, int x16, int y16, int rounder)
translational global motion compensation.
Definition: dsputil.h:212
qpel_mc_func put_2tap_qpel_pixels_tab[4][16]
Definition: dsputil.h:327
void(* add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale)
Definition: dsputil.h:435
static void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:676
int(* pix_norm1)(uint8_t *pix, int line_size)
Definition: dsputil.h:221
#define s1
Definition: regdef.h:38
int(* ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2, int size)
Definition: dsputil.h:246
short DCTELEM
Definition: dsputil.h:39
me_cmp_func w97[6]
Definition: dsputil.h:235
void(* h263_v_loop_filter)(uint8_t *src, int stride, int qscale)
Definition: dsputil.h:346
struct ScanTable ScanTable
Scantable.
void ff_fdct248_islow_8(DCTELEM *data)
void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride)
void(* vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len)
Definition: dsputil.h:354
void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable)
Definition: dsputil.c:122
uint8_t ff_cropTbl[256+2 *MAX_NEG_CROP]
Definition: dsputil.c:41
static uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
Definition: dsputil.h:524
static const uint16_t scale[4]
qpel_mc_func put_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:324
void(* put_no_rnd_pixels_l2[2])(uint8_t *block, const uint8_t *a, const uint8_t *b, int line_size, int h)
Definition: dsputil.h:297
int(* sum_abs_dctelem)(DCTELEM *block)
Definition: dsputil.h:208
void(* clear_blocks)(DCTELEM *blocks)
Definition: dsputil.h:219
void(* h261_loop_filter)(uint8_t *src, int stride)
Definition: dsputil.h:349
me_cmp_func rd[6]
Definition: dsputil.h:230
int height
Definition: gxfenc.c:72
tpel_mc_func put_tpel_pixels_tab[11]
Thirdpel motion compensation with rounding (a+b+1)>>1.
Definition: dsputil.h:309
void(* vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min, int32_t max, unsigned int len)
Clip each element in an array of int32_t to a given minimum and maximum value.
Definition: dsputil.h:486
static void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:665
void(* diff_bytes)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w)
Definition: dsputil.h:334
qpel_mc_func avg_h264_qpel_pixels_tab[4][16]
Definition: dsputil.h:325
void(* put_signed_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size)
Definition: dsputil.h:204
const uint8_t ff_zigzag_direct[64]
Definition: dsputil.c:59
void ff_jpeg_fdct_islow_8(DCTELEM *data)
void(* fdct)(DCTELEM *block)
Definition: dsputil.h:394
#define BYTE_VEC64(c)
Definition: dsputil.h:517
void(* gmc)(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)
global motion compensation.
Definition: dsputil.h:216
void ff_dsputil_init_sh4(DSPContext *c, AVCodecContext *avctx)
Definition: dsputil_sh4.c:92
static int16_t basis[64][64]
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)
Definition: dsputil.c:477
me_cmp_func sad[6]
Definition: dsputil.h:224
int32_t(* scalarproduct_and_madd_int16)(int16_t *v1, const int16_t *v2, const int16_t *v3, int len, int mul)
Calculate scalar product of v1 and v2, and v1[i] += v3[i] * mul.
Definition: dsputil.h:457
void(* tpel_mc_func)(uint8_t *block, const uint8_t *pixels, int line_size, int w, int h)
Definition: dsputil.h:143
float(* scalarproduct_float)(const float *v1, const float *v2, int len)
Calculate the scalar product of two vectors of floats.
Definition: dsputil.h:367
me_cmp_func hadamard8_diff[6]
Definition: dsputil.h:226
void(* butterflies_float_interleave)(float *dst, const float *src0, const float *src1, int len)
Calculate the sum and difference of two vectors of floats and interleave results into a separate outp...
Definition: dsputil.h:390
int len
void ff_fdct_mmxext(DCTELEM *block)
void(* add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size)
Definition: dsputil.h:207
me_cmp_func quant_psnr[6]
Definition: dsputil.h:228
static void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:653
void ff_fdct_ifast248(DCTELEM *data)
Definition: jfdctfst.c:274
qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]
Definition: dsputil.h:314
struct DSPContext DSPContext
DSPContext.
me_cmp_func me_sub_cmp[6]
Definition: dsputil.h:241
void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height)
Definition: imgconvert.c:225
static const uint16_t rounder[4]
#define restrict
Definition: config.h:8
float min
void ff_init_scantable_permutation(uint8_t *idct_permutation, int idct_permutation_type)
Definition: dsputil.c:143
static void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
Definition: dsputil.h:620
void ff_wmv2_idct_c(DCTELEM *data)
Definition: dsputil.c:2585
me_cmp_func pix_abs[2][4]
Definition: dsputil.h:330
void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride)
DSPContext.
Definition: dsputil.h:194
void(* op_fill_func)(uint8_t *block, uint8_t value, int line_size, int h)
Definition: dsputil.h:147