FFmpeg  1.2.4
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 FFmpeg.
7  *
8  * FFmpeg 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  * FFmpeg 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 FFmpeg; 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 
39 /* encoding scans */
40 extern const uint8_t ff_alternate_horizontal_scan[64];
41 extern const uint8_t ff_alternate_vertical_scan[64];
42 extern const uint8_t ff_zigzag_direct[64];
43 extern const uint8_t ff_zigzag248_direct[64];
44 
45 /* pixel operations */
46 #define MAX_NEG_CROP 1024
47 
48 /* temporary */
49 extern uint32_t ff_squareTbl[512];
50 extern const uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
51 
52 #define PUTAVG_PIXELS(depth)\
53 void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
54 void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
55 void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
56 void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);
57 
58 PUTAVG_PIXELS( 8)
59 PUTAVG_PIXELS( 9)
60 PUTAVG_PIXELS(10)
61 PUTAVG_PIXELS(12)
62 PUTAVG_PIXELS(14)
63 
64 #define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
65 #define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
66 #define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
67 #define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c
68 
69 /* RV40 functions */
74 
75 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
76  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
77 
78 /* minimum alignment rules ;)
79 If you notice errors in the align stuff, need more alignment for some ASM code
80 for some CPU or need to use a function with less aligned data then send a mail
81 to the ffmpeg-devel mailing list, ...
82 
83 !warning These alignments might not match reality, (missing attribute((align))
84 stuff somewhere possible).
85 I (Michael) did not check them, these are just the alignments which I think
86 could be reached easily ...
87 
88 !future video codecs might need functions with less strict alignment
89 */
90 
91 /* add and put pixel (decoding) */
92 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
93 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
94 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, ptrdiff_t line_size, int h);
95 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);
96 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
97 
98 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
99 
100 #define DEF_OLD_QPEL(name)\
101 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
102 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
103 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
104 
105 DEF_OLD_QPEL(qpel16_mc11_old_c)
106 DEF_OLD_QPEL(qpel16_mc31_old_c)
107 DEF_OLD_QPEL(qpel16_mc12_old_c)
108 DEF_OLD_QPEL(qpel16_mc32_old_c)
109 DEF_OLD_QPEL(qpel16_mc13_old_c)
110 DEF_OLD_QPEL(qpel16_mc33_old_c)
111 DEF_OLD_QPEL(qpel8_mc11_old_c)
112 DEF_OLD_QPEL(qpel8_mc31_old_c)
113 DEF_OLD_QPEL(qpel8_mc12_old_c)
114 DEF_OLD_QPEL(qpel8_mc32_old_c)
115 DEF_OLD_QPEL(qpel8_mc13_old_c)
116 DEF_OLD_QPEL(qpel8_mc33_old_c)
117 
118 #define CALL_2X_PIXELS(a, b, n)\
119 static void a(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h){\
120  b(block , pixels , line_size, h);\
121  b(block+n, pixels+n, line_size, h);\
122 }
123 
124 /* motion estimation */
125 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
126 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
127 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))*/;
128 
132 typedef struct ScanTable{
136 } ScanTable;
137 
138 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
139 void ff_init_scantable_permutation(uint8_t *idct_permutation,
140  int idct_permutation_type);
141 
145 typedef struct DSPContext {
149  int dct_bits;
150 
151  /* pixel ops : interface with DCT */
152  void (*get_pixels)(int16_t *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
153  void (*diff_pixels)(int16_t *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
154  void (*put_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
155  void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
156  void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
157  void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
158  int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
162  void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
166  void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
167  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
168  void (*clear_block)(int16_t *block/*align 16*/);
169  void (*clear_blocks)(int16_t *blocks/*align 16*/);
170  int (*pix_sum)(uint8_t * pix, int line_size);
171  int (*pix_norm1)(uint8_t * pix, int line_size);
172 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
173 
174  me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
188 
193  me_cmp_func ildct_cmp[6]; //only width 16 used
194  me_cmp_func frame_skip_cmp[6]; //only width 8 used
195 
196  int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
197  int size);
198 
210 
222 
234 
246 
257  tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
258  tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
259 
264 
266 
267  /* huffyuv specific */
268  void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
269  void (*diff_bytes)(uint8_t *dst/*align 16*/, const uint8_t *src1/*align 16*/, const uint8_t *src2/*align 1*/,int w);
274  void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
275  void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
276  int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
277  void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
278  /* this might write to dst[w] */
279  void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
280  void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
281 
282  void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
283  void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
284 
286 
287  /* assume len is a multiple of 8, and arrays are 16-byte aligned */
288  void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
289 
290  /* (I)DCT */
291  void (*fdct)(int16_t *block/* align 16*/);
292  void (*fdct248)(int16_t *block/* align 16*/);
293 
294  /* IDCT really*/
295  void (*idct)(int16_t *block/* align 16*/);
296 
302  void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
303 
308  void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
309 
324 #define FF_NO_IDCT_PERM 1
325 #define FF_LIBMPEG2_IDCT_PERM 2
326 #define FF_SIMPLE_IDCT_PERM 3
327 #define FF_TRANSPOSE_IDCT_PERM 4
328 #define FF_PARTTRANS_IDCT_PERM 5
329 #define FF_SSE2_IDCT_PERM 6
330 
331  int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
332  void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
333 #define BASIS_SHIFT 16
334 #define RECON_SHIFT 6
335 
336  void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
337 #define EDGE_WIDTH 16
338 #define EDGE_TOP 1
339 #define EDGE_BOTTOM 2
340 
341  void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
342 
347  int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
348  /* ape functions */
354  int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
355 
367  void (*apply_window_int16)(int16_t *output, const int16_t *input,
368  const int16_t *window, unsigned int len);
369 
384  int32_t max, unsigned int len);
385 
387 } DSPContext;
388 
389 void ff_dsputil_static_init(void);
390 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
392 
393 int ff_check_alignment(void);
394 
395 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
396 
404 
406 
407 #endif /* AVCODEC_DSPUTIL_H */