Libav
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 #include "rnd_avg.h"
36 
37 /* encoding scans */
38 extern const uint8_t ff_alternate_horizontal_scan[64];
39 extern const uint8_t ff_alternate_vertical_scan[64];
40 extern const uint8_t ff_zigzag_direct[64];
41 extern const uint8_t ff_zigzag248_direct[64];
42 
43 /* pixel operations */
44 #define MAX_NEG_CROP 1024
45 
46 /* temporary */
47 extern uint32_t ff_squareTbl[512];
48 extern const uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
49 
50 void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
51 void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
52 void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
53 void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
54 
55 /* RV40 functions */
56 void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
57 void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
58 void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
59 void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
60 
61 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
62  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
63 
64 /* minimum alignment rules ;)
65 If you notice errors in the align stuff, need more alignment for some ASM code
66 for some CPU or need to use a function with less aligned data then send a mail
67 to the libav-devel mailing list, ...
68 
69 !warning These alignments might not match reality, (missing attribute((align))
70 stuff somewhere possible).
71 I (Michael) did not check them, these are just the alignments which I think
72 could be reached easily ...
73 
74 !future video codecs might need functions with less strict alignment
75 */
76 
77 /* add and put pixel (decoding) */
78 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
79 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
80 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);
81 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);
82 
83 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
84 
85 #define DEF_OLD_QPEL(name)\
86 void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);\
87 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);\
88 void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);
89 
90 DEF_OLD_QPEL(qpel16_mc11_old_c)
91 DEF_OLD_QPEL(qpel16_mc31_old_c)
92 DEF_OLD_QPEL(qpel16_mc12_old_c)
93 DEF_OLD_QPEL(qpel16_mc32_old_c)
94 DEF_OLD_QPEL(qpel16_mc13_old_c)
95 DEF_OLD_QPEL(qpel16_mc33_old_c)
96 DEF_OLD_QPEL(qpel8_mc11_old_c)
97 DEF_OLD_QPEL(qpel8_mc31_old_c)
98 DEF_OLD_QPEL(qpel8_mc12_old_c)
99 DEF_OLD_QPEL(qpel8_mc32_old_c)
100 DEF_OLD_QPEL(qpel8_mc13_old_c)
101 DEF_OLD_QPEL(qpel8_mc33_old_c)
102 
103 /* motion estimation */
104 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
105 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
106 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))*/;
107 
111 typedef struct ScanTable{
113  uint8_t permutated[64];
114  uint8_t raster_end[64];
115 } ScanTable;
116 
117 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
118 void ff_init_scantable_permutation(uint8_t *idct_permutation,
119  int idct_permutation_type);
120 
124 typedef struct DSPContext {
125  /* pixel ops : interface with DCT */
126  void (*get_pixels)(int16_t *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
127  void (*diff_pixels)(int16_t *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
128  void (*put_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
129  void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
130  void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
131  void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
132  int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
136  void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
140  void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
141  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
142  void (*clear_block)(int16_t *block/*align 16*/);
143  void (*clear_blocks)(int16_t *blocks/*align 16*/);
144  int (*pix_sum)(uint8_t * pix, int line_size);
145  int (*pix_norm1)(uint8_t * pix, int line_size);
146 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
147 
148  me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
160 
165  me_cmp_func ildct_cmp[6]; //only width 16 used
166  me_cmp_func frame_skip_cmp[6]; //only width 8 used
167 
168  int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
169  int size);
170 
181  tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
182  tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
183 
188 
190 
191  /* huffyuv specific */
192  void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
193  void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
198  void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
199  void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
200  int (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
201  void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
202  void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
203  void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
204 
205  /* assume len is a multiple of 8, and arrays are 16-byte aligned */
206  void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
207 
208  /* (I)DCT */
209  void (*fdct)(int16_t *block/* align 16*/);
210  void (*fdct248)(int16_t *block/* align 16*/);
211 
212  /* IDCT really*/
213  void (*idct)(int16_t *block/* align 16*/);
214 
220  void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
221 
226  void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);
227 
242 #define FF_NO_IDCT_PERM 1
243 #define FF_LIBMPEG2_IDCT_PERM 2
244 #define FF_SIMPLE_IDCT_PERM 3
245 #define FF_TRANSPOSE_IDCT_PERM 4
246 #define FF_PARTTRANS_IDCT_PERM 5
247 #define FF_SSE2_IDCT_PERM 6
248 
249  int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
250  void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
251 #define BASIS_SHIFT 16
252 #define RECON_SHIFT 6
253 
254  void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
255 #define EDGE_WIDTH 16
256 #define EDGE_TOP 1
257 #define EDGE_BOTTOM 2
258 
259  void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
260 
265  int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
266  /* ape functions */
272  int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
273 
288  int32_t max, unsigned int len);
289 
291 } DSPContext;
292 
293 void ff_dsputil_static_init(void);
294 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
295 
296 int ff_check_alignment(void);
297 
298 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
299 
306 
307 #endif /* AVCODEC_DSPUTIL_H */