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
|
/*
* DSP utils
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard.
*
* This library 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 of the License, or (at your option) any later version.
*
* This library 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 this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#ifndef DSPUTIL_H
#define DSPUTIL_H
#include "common.h"
#include "avcodec.h"
//#define DEBUG
/* dct code */
typedef short DCTELEM;
void fdct_ifast (DCTELEM *data);
void ff_jpeg_fdct_islow (DCTELEM *data);
void j_rev_dct (DCTELEM *data);
void ff_fdct_mmx(DCTELEM *block);
/* encoding scans */
extern const UINT8 ff_alternate_horizontal_scan[64];
extern const UINT8 ff_alternate_vertical_scan[64];
extern const UINT8 ff_zigzag_direct[64];
/* pixel operations */
#define MAX_NEG_CROP 384
/* temporary */
extern UINT32 squareTbl[512];
extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];
void dsputil_init(void);
/* minimum alignment rules ;)
if u 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-dev list, ...
!warning these alignments might not match reallity, (missing attribute((align)) stuff somewhere possible)
i (michael) didnt check them, these are just the alignents which i think could be reached easily ...
!future video codecs might need functions with less strict alignment
*/
/* pixel ops : interface with DCT */
extern void (*get_pixels)(DCTELEM *block/*align 16*/, const UINT8 *pixels/*align 8*/, int line_size);
extern void (*diff_pixels)(DCTELEM *block/*align 16*/, const UINT8 *s1/*align 8*/, const UINT8 *s2/*align 8*/, int stride);
extern void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size);
extern void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, UINT8 *pixels/*align 8*/, int line_size);
extern void (*ff_gmc1)(UINT8 *dst/*align 8*/, UINT8 *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
extern void (*ff_gmc )(UINT8 *dst/*align 8*/, UINT8 *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);
extern void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
extern int (*pix_sum)(UINT8 * pix, int line_size);
extern int (*pix_norm1)(UINT8 * pix, int line_size);
void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size);
void diff_pixels_c(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride);
void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size);
void add_pixels_clamped_c(const DCTELEM *block, UINT8 *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
typedef void (*op_pixels_func)(UINT8 *block/*align width (8 or 16)*/, const UINT8 *pixels/*align 1*/, int line_size, int h);
typedef void (*qpel_mc_func)(UINT8 *dst/*align width (8 or 16)*/, UINT8 *src/*align 1*/, int stride);
extern op_pixels_func put_pixels_tab[2][4];
extern op_pixels_func avg_pixels_tab[2][4];
extern op_pixels_func put_no_rnd_pixels_tab[2][4];
extern op_pixels_func avg_no_rnd_pixels_tab[2][4];
extern qpel_mc_func put_qpel_pixels_tab[2][16];
extern qpel_mc_func avg_qpel_pixels_tab[2][16];
extern qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
extern qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
#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 */
typedef int (*op_pixels_abs_func)(UINT8 *blk1/*align width (8 or 16)*/, UINT8 *blk2/*align 1*/, int line_size);
extern op_pixels_abs_func pix_abs16x16;
extern op_pixels_abs_func pix_abs16x16_x2;
extern op_pixels_abs_func pix_abs16x16_y2;
extern op_pixels_abs_func pix_abs16x16_xy2;
extern op_pixels_abs_func pix_abs8x8;
extern op_pixels_abs_func pix_abs8x8_x2;
extern op_pixels_abs_func pix_abs8x8_y2;
extern op_pixels_abs_func pix_abs8x8_xy2;
int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx);
int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx);
int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx);
int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx);
/**
* permute block according to permuatation.
* @param last last non zero element in scantable order
*/
void ff_block_permute(INT16 *block, UINT8 *permutation, const UINT8 *scantable, int last);
#if defined(HAVE_MMX)
#define MM_MMX 0x0001 /* standard MMX */
#define MM_3DNOW 0x0004 /* AMD 3DNOW */
#define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */
#define MM_SSE 0x0008 /* SSE functions */
#define MM_SSE2 0x0010 /* PIV SSE2 functions */
extern int mm_flags;
int mm_support(void);
static inline void emms(void)
{
__asm __volatile ("emms;":::"memory");
}
#define emms_c() \
{\
if (mm_flags & MM_MMX)\
emms();\
}
#define __align8 __attribute__ ((aligned (8)))
void dsputil_init_mmx(void);
void dsputil_set_bit_exact_mmx(void);
#elif defined(ARCH_ARMV4L)
#define emms_c()
/* This is to use 4 bytes read to the IDCT pointers for some 'zero'
line ptimizations */
#define __align8 __attribute__ ((aligned (4)))
void dsputil_init_armv4l(void);
#elif defined(HAVE_MLIB)
#define emms_c()
/* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */
#define __align8 __attribute__ ((aligned (8)))
void dsputil_init_mlib(void);
#elif defined(ARCH_ALPHA)
#define emms_c()
#define __align8 __attribute__ ((aligned (8)))
void dsputil_init_alpha(void);
#elif defined(ARCH_POWERPC)
#define emms_c()
#define __align8 __attribute__ ((aligned (16)))
void dsputil_init_ppc(void);
#elif defined(HAVE_MMI)
#define emms_c()
#define __align8 __attribute__ ((aligned (16)))
void dsputil_init_mmi(void);
#else
#define emms_c()
#define __align8
#endif
#ifdef __GNUC__
struct unaligned_64 { uint64_t l; } __attribute__((packed));
struct unaligned_32 { uint32_t l; } __attribute__((packed));
#define LD32(a) (((const struct unaligned_32 *) (a))->l)
#define LD64(a) (((const struct unaligned_64 *) (a))->l)
#define ST32(a, b) (((struct unaligned_32 *) (a))->l) = (b)
#else /* __GNUC__ */
#define LD32(a) (*((uint32_t*)(a)))
#define LD64(a) (*((uint64_t*)(a)))
#define ST32(a, b) *((uint32_t*)(a)) = (b)
#endif /* !__GNUC__ */
/* PSNR */
void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3],
int orig_linesize[3], int coded_linesize,
AVCodecContext *avctx);
/* FFT computation */
/* NOTE: soon integer code will be added, so you must use the
FFTSample type */
typedef float FFTSample;
typedef struct FFTComplex {
FFTSample re, im;
} FFTComplex;
typedef struct FFTContext {
int nbits;
int inverse;
uint16_t *revtab;
FFTComplex *exptab;
FFTComplex *exptab1; /* only used by SSE code */
void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
} FFTContext;
int fft_init(FFTContext *s, int nbits, int inverse);
void fft_permute(FFTContext *s, FFTComplex *z);
void fft_calc_c(FFTContext *s, FFTComplex *z);
void fft_calc_sse(FFTContext *s, FFTComplex *z);
static inline void fft_calc(FFTContext *s, FFTComplex *z)
{
s->fft_calc(s, z);
}
void fft_end(FFTContext *s);
/* MDCT computation */
typedef struct MDCTContext {
int n; /* size of MDCT (i.e. number of input data * 2) */
int nbits; /* n = 2^nbits */
/* pre/post rotation tables */
FFTSample *tcos;
FFTSample *tsin;
FFTContext fft;
} MDCTContext;
int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
void ff_imdct_calc(MDCTContext *s, FFTSample *output,
const FFTSample *input, FFTSample *tmp);
void ff_mdct_calc(MDCTContext *s, FFTSample *out,
const FFTSample *input, FFTSample *tmp);
void ff_mdct_end(MDCTContext *s);
#if defined(__FreeBSD__) || (__bsdi__)
/* XXX: add ISOC specific test to avoid specific BSD testing. */
/* better than nothing implementation. */
static inline long int lrintf(float x)
{
return (int)(rint(x));
}
#endif
#endif
|