Mercurial > libavcodec.hg
annotate dsputil.h @ 294:944632089814 libavcodec
4MV motion estimation (not finished yet)
SAD functions rewritten (8x8 support & MMX2 optimizations)
HQ inter/intra decission
msmpeg4 encoding bugfix (MV where too long)
author | michaelni |
---|---|
date | Wed, 27 Mar 2002 21:25:22 +0000 |
parents | db20b987c32d |
children | c1a8a1b4a24b |
rev | line source |
---|---|
0 | 1 #ifndef DSPUTIL_H |
2 #define DSPUTIL_H | |
3 | |
4 #include "common.h" | |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
5 #include "avcodec.h" |
0 | 6 |
255 | 7 //#define DEBUG |
0 | 8 /* dct code */ |
9 typedef short DCTELEM; | |
10 | |
11 void jpeg_fdct_ifast (DCTELEM *data); | |
12 | |
13 void j_rev_dct (DCTELEM *data); | |
14 | |
15 void fdct_mmx(DCTELEM *block); | |
16 | |
17 void (*av_fdct)(DCTELEM *block); | |
18 | |
34 | 19 /* encoding scans */ |
20 extern UINT8 ff_alternate_horizontal_scan[64]; | |
21 extern UINT8 ff_alternate_vertical_scan[64]; | |
22 extern UINT8 zigzag_direct[64]; | |
23 | |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
24 /* permutation table */ |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
25 extern UINT8 permutation[64]; |
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
26 |
0 | 27 /* pixel operations */ |
28 #define MAX_NEG_CROP 384 | |
29 | |
30 /* temporary */ | |
31 extern UINT32 squareTbl[512]; | |
50 | 32 extern UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; |
0 | 33 |
34 void dsputil_init(void); | |
35 | |
36 /* pixel ops : interface with DCT */ | |
37 | |
19
82d4c9be9873
MMX/MMXEXT iDCT support, using external functions currently defined in libmpeg2
arpi_esp
parents:
6
diff
changeset
|
38 extern void (*ff_idct)(DCTELEM *block); |
0 | 39 extern void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size); |
40 extern void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
41 extern void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size); | |
255 | 42 extern void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder); |
43 | |
0 | 44 |
45 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
46 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
47 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
48 | |
49 /* add and put pixel (decoding) */ | |
50 typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h); | |
255 | 51 typedef void (*qpel_mc_func)(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my); |
0 | 52 |
53 extern op_pixels_func put_pixels_tab[4]; | |
54 extern op_pixels_func avg_pixels_tab[4]; | |
55 extern op_pixels_func put_no_rnd_pixels_tab[4]; | |
56 extern op_pixels_func avg_no_rnd_pixels_tab[4]; | |
255 | 57 extern qpel_mc_func qpel_mc_rnd_tab[16]; |
58 extern qpel_mc_func qpel_mc_no_rnd_tab[16]; | |
59 | |
0 | 60 |
61 /* sub pixel (encoding) */ | |
62 extern void (*sub_pixels_tab[4])(DCTELEM *block, const UINT8 *pixels, int line_size, int h); | |
63 | |
64 #define sub_pixels_2(block, pixels, line_size, dxy) \ | |
65 sub_pixels_tab[dxy](block, pixels, line_size, 8) | |
66 | |
67 /* motion estimation */ | |
68 | |
294 | 69 typedef int (*op_pixels_abs_func)(UINT8 *blk1, UINT8 *blk2, int line_size); |
0 | 70 |
71 extern op_pixels_abs_func pix_abs16x16; | |
72 extern op_pixels_abs_func pix_abs16x16_x2; | |
73 extern op_pixels_abs_func pix_abs16x16_y2; | |
74 extern op_pixels_abs_func pix_abs16x16_xy2; | |
294 | 75 extern op_pixels_abs_func pix_abs8x8; |
76 extern op_pixels_abs_func pix_abs8x8_x2; | |
77 extern op_pixels_abs_func pix_abs8x8_y2; | |
78 extern op_pixels_abs_func pix_abs8x8_xy2; | |
0 | 79 |
294 | 80 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
81 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
82 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
83 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
0 | 84 |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
85 static inline int block_permute_op(int j) |
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
86 { |
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
87 return permutation[j]; |
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
88 } |
34 | 89 |
90 void block_permute(INT16 *block); | |
91 | |
62 | 92 #if defined(HAVE_MMX) |
0 | 93 |
94 #define MM_MMX 0x0001 /* standard MMX */ | |
95 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
96 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
97 #define MM_SSE 0x0008 /* SSE functions */ | |
98 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
99 | |
100 extern int mm_flags; | |
101 | |
102 int mm_support(void); | |
103 | |
104 static inline void emms(void) | |
105 { | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
106 __asm __volatile ("emms;":::"memory"); |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
107 } |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
108 |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
109 #define emms_c() \ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
110 {\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
111 if (mm_flags & MM_MMX)\ |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
112 emms();\ |
0 | 113 } |
114 | |
115 #define __align8 __attribute__ ((aligned (8))) | |
116 | |
117 void dsputil_init_mmx(void); | |
118 | |
62 | 119 #elif defined(ARCH_ARMV4L) |
120 | |
121 #define emms_c() | |
122 | |
123 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
124 line ptimizations */ | |
125 #define __align8 __attribute__ ((aligned (4))) | |
126 | |
127 void dsputil_init_armv4l(void); | |
128 | |
88 | 129 #elif defined(HAVE_MLIB) |
130 | |
131 #define emms_c() | |
132 | |
133 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
134 #define __align8 __attribute__ ((aligned (8))) | |
135 | |
136 void dsputil_init_mlib(void); | |
137 | |
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
138 #elif defined(ARCH_ALPHA) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
139 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
140 #define emms_c() |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
141 #define __align8 __attribute__ ((aligned (8))) |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
142 |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
143 void dsputil_init_alpha(void); |
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
144 |
0 | 145 #else |
146 | |
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
147 #define emms_c() |
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
148 |
0 | 149 #define __align8 |
150 | |
151 #endif | |
152 | |
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
153 /* PSNR */ |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
154 void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3], |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
155 int orig_linesize[3], int coded_linesize, |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
156 AVCodecContext *avctx); |
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
157 |
0 | 158 #endif |