Mercurial > libavcodec.hg
annotate dsputil.h @ 296:c1a8a1b4a24b libavcodec
sizeof(s->block) isnt 64*6*2 anymore bugfix
mpeg12 decoding optimization
| author | michaelni |
|---|---|
| date | Thu, 28 Mar 2002 13:41:04 +0000 |
| parents | 944632089814 |
| children | 9c6f056f0e41 |
| 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); |
| 296 | 43 extern void (*clear_blocks)(DCTELEM *blocks); |
| 255 | 44 |
| 0 | 45 |
| 46 void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size); | |
| 47 void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
| 48 void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size); | |
| 296 | 49 void clear_blocks_c(DCTELEM *blocks); |
| 0 | 50 |
| 51 /* add and put pixel (decoding) */ | |
| 52 typedef void (*op_pixels_func)(UINT8 *block, const UINT8 *pixels, int line_size, int h); | |
| 255 | 53 typedef void (*qpel_mc_func)(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int mx, int my); |
| 0 | 54 |
| 55 extern op_pixels_func put_pixels_tab[4]; | |
| 56 extern op_pixels_func avg_pixels_tab[4]; | |
| 57 extern op_pixels_func put_no_rnd_pixels_tab[4]; | |
| 58 extern op_pixels_func avg_no_rnd_pixels_tab[4]; | |
| 255 | 59 extern qpel_mc_func qpel_mc_rnd_tab[16]; |
| 60 extern qpel_mc_func qpel_mc_no_rnd_tab[16]; | |
| 61 | |
| 0 | 62 |
| 63 /* sub pixel (encoding) */ | |
| 64 extern void (*sub_pixels_tab[4])(DCTELEM *block, const UINT8 *pixels, int line_size, int h); | |
| 65 | |
| 66 #define sub_pixels_2(block, pixels, line_size, dxy) \ | |
| 67 sub_pixels_tab[dxy](block, pixels, line_size, 8) | |
| 68 | |
| 69 /* motion estimation */ | |
| 70 | |
| 294 | 71 typedef int (*op_pixels_abs_func)(UINT8 *blk1, UINT8 *blk2, int line_size); |
| 0 | 72 |
| 73 extern op_pixels_abs_func pix_abs16x16; | |
| 74 extern op_pixels_abs_func pix_abs16x16_x2; | |
| 75 extern op_pixels_abs_func pix_abs16x16_y2; | |
| 76 extern op_pixels_abs_func pix_abs16x16_xy2; | |
| 294 | 77 extern op_pixels_abs_func pix_abs8x8; |
| 78 extern op_pixels_abs_func pix_abs8x8_x2; | |
| 79 extern op_pixels_abs_func pix_abs8x8_y2; | |
| 80 extern op_pixels_abs_func pix_abs8x8_xy2; | |
| 0 | 81 |
| 294 | 82 int pix_abs16x16_c(UINT8 *blk1, UINT8 *blk2, int lx); |
| 83 int pix_abs16x16_x2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
| 84 int pix_abs16x16_y2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
| 85 int pix_abs16x16_xy2_c(UINT8 *blk1, UINT8 *blk2, int lx); | |
| 0 | 86 |
|
174
ac5075a55488
new IDCT code by Michael Niedermayer (michaelni@gmx.at) - #define SIMPLE_IDCT to enable
arpi_esp
parents:
88
diff
changeset
|
87 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
|
88 { |
|
190
9e0e56869d05
fix for non-mmx runtimedetect encoding bugs - patch by Michael Niedermayer <michaelni@gmx.at>
uid46427
parents:
174
diff
changeset
|
89 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
|
90 } |
| 34 | 91 |
| 92 void block_permute(INT16 *block); | |
| 93 | |
| 62 | 94 #if defined(HAVE_MMX) |
| 0 | 95 |
| 96 #define MM_MMX 0x0001 /* standard MMX */ | |
| 97 #define MM_3DNOW 0x0004 /* AMD 3DNOW */ | |
| 98 #define MM_MMXEXT 0x0002 /* SSE integer functions or AMD MMX ext */ | |
| 99 #define MM_SSE 0x0008 /* SSE functions */ | |
| 100 #define MM_SSE2 0x0010 /* PIV SSE2 functions */ | |
| 101 | |
| 102 extern int mm_flags; | |
| 103 | |
| 104 int mm_support(void); | |
| 105 | |
| 106 static inline void emms(void) | |
| 107 { | |
|
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
108 __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
|
109 } |
|
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 #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
|
112 {\ |
|
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
113 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
|
114 emms();\ |
| 0 | 115 } |
| 116 | |
| 117 #define __align8 __attribute__ ((aligned (8))) | |
| 118 | |
| 119 void dsputil_init_mmx(void); | |
| 120 | |
| 62 | 121 #elif defined(ARCH_ARMV4L) |
| 122 | |
| 123 #define emms_c() | |
| 124 | |
| 125 /* This is to use 4 bytes read to the IDCT pointers for some 'zero' | |
| 126 line ptimizations */ | |
| 127 #define __align8 __attribute__ ((aligned (4))) | |
| 128 | |
| 129 void dsputil_init_armv4l(void); | |
| 130 | |
| 88 | 131 #elif defined(HAVE_MLIB) |
| 132 | |
| 133 #define emms_c() | |
| 134 | |
| 135 /* SPARC/VIS IDCT needs 8-byte aligned DCT blocks */ | |
| 136 #define __align8 __attribute__ ((aligned (8))) | |
| 137 | |
| 138 void dsputil_init_mlib(void); | |
| 139 | |
|
214
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
140 #elif defined(ARCH_ALPHA) |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
141 |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
142 #define emms_c() |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
143 #define __align8 __attribute__ ((aligned (8))) |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
144 |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
145 void dsputil_init_alpha(void); |
|
73df666cacc7
Alpha optimizations by Falk Hueffner <falk.hueffner@student.uni-tuebingen.de>
nickols_k
parents:
190
diff
changeset
|
146 |
| 0 | 147 #else |
| 148 | |
|
6
ec4642daa6fe
added emms_c() macro which should can used in c code in both mmx/non mmx cases
glantau
parents:
2
diff
changeset
|
149 #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
|
150 |
| 0 | 151 #define __align8 |
| 152 | |
| 153 #endif | |
| 154 | |
|
252
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
155 /* PSNR */ |
|
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
156 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
|
157 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
|
158 AVCodecContext *avctx); |
|
ddb1a0e94cf4
- Added PSNR feature to libavcodec and ffmpeg. By now just Y PSNR until I'm
pulento
parents:
214
diff
changeset
|
159 |
| 0 | 160 #endif |
