Add forgotten cleanup function in atrac1.
[libav.git] / libavcodec / dsputil.h
CommitLineData
ff4ec49e
FB
1/*
2 * DSP utils
406792e7 3 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
8f2ab833 4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
ff4ec49e 5 *
b78e7197
DB
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
ff4ec49e
FB
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
b78e7197 11 * version 2.1 of the License, or (at your option) any later version.
ff4ec49e 12 *
b78e7197 13 * FFmpeg is distributed in the hope that it will be useful,
ff4ec49e
FB
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
b78e7197 19 * License along with FFmpeg; if not, write to the Free Software
5509bffa 20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
ff4ec49e 21 */
24641185
MN
22
23/**
bad5537e 24 * @file libavcodec/dsputil.h
983e3246 25 * DSP utils.
5755c27f
MN
26 * note, many functions in here may use MMX which trashes the FPU state, it is
27 * absolutely necessary to call emms_c() between dsp & float/double code
24641185
MN
28 */
29
98790382
SS
30#ifndef AVCODEC_DSPUTIL_H
31#define AVCODEC_DSPUTIL_H
de6d9b64 32
6a5d31ac 33#include "libavutil/intreadwrite.h"
43f1708f 34#include "avcodec.h"
de6d9b64 35
24641185 36
44eb4951 37//#define DEBUG
de6d9b64
FB
38/* dct code */
39typedef short DCTELEM;
059715a4 40typedef int DWTELEM;
d593e329 41typedef short IDWTELEM;
de6d9b64 42
03c94ede 43void fdct_ifast (DCTELEM *data);
48b1f800 44void fdct_ifast248 (DCTELEM *data);
28db7fce 45void ff_jpeg_fdct_islow (DCTELEM *data);
10acc479 46void ff_fdct248_islow (DCTELEM *data);
de6d9b64
FB
47
48void j_rev_dct (DCTELEM *data);
178fcca8 49void j_rev_dct4 (DCTELEM *data);
9ca358b9 50void j_rev_dct2 (DCTELEM *data);
1aa8c57b 51void j_rev_dct1 (DCTELEM *data);
9abc7e0f 52void ff_wmv2_idct_c(DCTELEM *data);
de6d9b64 53
3f09f52a 54void ff_fdct_mmx(DCTELEM *block);
cf3bf5bb 55void ff_fdct_mmx2(DCTELEM *block);
8fd19ab2 56void ff_fdct_sse2(DCTELEM *block);
de6d9b64 57
43efd19a 58void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
0fa8158d 59void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride);
ef9d1d15
LM
60void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
61void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
0fa8158d
MN
62void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
63void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
ac223859
MN
64void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
65void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
66void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
67void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
0fa8158d 68
eb4825b5
LM
69void ff_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
70 const float *src2, int src3, int blocksize, int step);
f27e1d64
LM
71void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
72 const float *win, float add_bias, int len);
560fa9bf 73void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
8a37920c 74void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels);
eb4825b5 75
e0eac44e 76/* encoding scans */
0c1a9eda
ZK
77extern const uint8_t ff_alternate_horizontal_scan[64];
78extern const uint8_t ff_alternate_vertical_scan[64];
79extern const uint8_t ff_zigzag_direct[64];
10acc479 80extern const uint8_t ff_zigzag248_direct[64];
5a240838 81
de6d9b64 82/* pixel operations */
f2e92ef2 83#define MAX_NEG_CROP 1024
de6d9b64
FB
84
85/* temporary */
1d503957 86extern uint32_t ff_squareTbl[512];
55fde95e 87extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
de6d9b64 88
44cb64ee 89/* VP3 DSP functions */
8b6103da
MN
90void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
91void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
92void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
de6d9b64 93
9971331d
DC
94void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
95void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
96
26f3ae82
AJ
97/* VP6 DSP functions */
98void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
99 const int16_t *h_weights, const int16_t *v_weights);
100
54009d42
MN
101/* 1/2^n downscaling functions from imgconvert.c */
102void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
103void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
104void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
105void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
703c8195
LM
106
107void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
108 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
54009d42 109
b7c27ee6 110/* minimum alignment rules ;)
7ce68923
DB
111If you notice errors in the align stuff, need more alignment for some ASM code
112for some CPU or need to use a function with less aligned data then send a mail
113to the ffmpeg-devel mailing list, ...
114
115!warning These alignments might not match reality, (missing attribute((align))
116stuff somewhere possible).
2cab6401 117I (Michael) did not check them, these are just the alignments which I think
7ce68923 118could be reached easily ...
de6d9b64 119
b7c27ee6
MN
120!future video codecs might need functions with less strict alignment
121*/
122
eb4b3dd3 123/*
0c1a9eda
ZK
124void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
125void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
126void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
127void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
649c00c9 128void clear_blocks_c(DCTELEM *blocks);
eb4b3dd3 129*/
de6d9b64
FB
130
131/* add and put pixel (decoding) */
b7c27ee6 132// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
7d67aa9b 133//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
0c1a9eda 134typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
669ac79c 135typedef 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);
0c1a9eda 136typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
0da71265 137typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
9f2d1b4f 138typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
e8b56208 139typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
b3184779 140
db794953 141#define DEF_OLD_QPEL(name)\
0c1a9eda
ZK
142void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
143void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
144void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
db794953
MN
145
146DEF_OLD_QPEL(qpel16_mc11_old_c)
147DEF_OLD_QPEL(qpel16_mc31_old_c)
148DEF_OLD_QPEL(qpel16_mc12_old_c)
149DEF_OLD_QPEL(qpel16_mc32_old_c)
150DEF_OLD_QPEL(qpel16_mc13_old_c)
151DEF_OLD_QPEL(qpel16_mc33_old_c)
152DEF_OLD_QPEL(qpel8_mc11_old_c)
153DEF_OLD_QPEL(qpel8_mc31_old_c)
154DEF_OLD_QPEL(qpel8_mc12_old_c)
155DEF_OLD_QPEL(qpel8_mc32_old_c)
156DEF_OLD_QPEL(qpel8_mc13_old_c)
157DEF_OLD_QPEL(qpel8_mc33_old_c)
b3184779
MN
158
159#define CALL_2X_PIXELS(a, b, n)\
160static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
161 b(block , pixels , line_size, h);\
162 b(block+n, pixels+n, line_size, h);\
163}
44eb4951 164
de6d9b64 165/* motion estimation */
7d67aa9b 166// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
755bfeab 167// although currently h<4 is not used as functions with width <8 are neither used nor implemented
bb198e19 168typedef 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))*/;
1457ab52 169
0da71265 170
059715a4
RE
171// for snow slices
172typedef struct slice_buffer_s slice_buffer;
173
24641185 174/**
4c79b95c
AJ
175 * Scantable.
176 */
177typedef struct ScanTable{
178 const uint8_t *scantable;
179 uint8_t permutated[64];
180 uint8_t raster_end[64];
b250f9c6 181#if ARCH_PPC
4c79b95c 182 /** Used by dct_quantize_altivec to find last-non-zero */
49c35f16 183 DECLARE_ALIGNED(16, uint8_t, inverse[64]);
4c79b95c
AJ
184#endif
185} ScanTable;
186
187void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
188
288a44fb
AJ
189void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize,
190 int block_w, int block_h,
191 int src_x, int src_y, int w, int h);
192
4c79b95c 193/**
24641185
MN
194 * DSPContext.
195 */
eb4b3dd3
ZK
196typedef struct DSPContext {
197 /* pixel ops : interface with DCT */
0c1a9eda
ZK
198 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
199 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
200 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
f9ed9d85 201 void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
0c1a9eda 202 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
36940eca
LM
203 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
204 void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
1edbfe19 205 int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
d518aebd
MN
206 /**
207 * translational global motion compensation.
208 */
0c1a9eda 209 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
d518aebd
MN
210 /**
211 * global motion compensation.
212 */
0c1a9eda 213 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
bb270c08 214 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
5fecfb7d 215 void (*clear_block)(DCTELEM *block/*align 16*/);
eb4b3dd3 216 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
0c1a9eda
ZK
217 int (*pix_sum)(uint8_t * pix, int line_size);
218 int (*pix_norm1)(uint8_t * pix, int line_size);
bb198e19 219// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
115329f1 220
3899eb2f
RS
221 me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
222 me_cmp_func sse[6];
223 me_cmp_func hadamard8_diff[6];
224 me_cmp_func dct_sad[6];
225 me_cmp_func quant_psnr[6];
226 me_cmp_func bit[6];
227 me_cmp_func rd[6];
228 me_cmp_func vsad[6];
229 me_cmp_func vsse[6];
230 me_cmp_func nsse[6];
231 me_cmp_func w53[6];
232 me_cmp_func w97[6];
233 me_cmp_func dct_max[6];
234 me_cmp_func dct264_sad[6];
235
236 me_cmp_func me_pre_cmp[6];
237 me_cmp_func me_cmp[6];
238 me_cmp_func me_sub_cmp[6];
239 me_cmp_func mb_cmp[6];
240 me_cmp_func ildct_cmp[6]; //only width 16 used
241 me_cmp_func frame_skip_cmp[6]; //only width 8 used
eb4b3dd3 242
a00177a9
MR
243 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
244 int size);
59006372 245
d518aebd
MN
246 /**
247 * Halfpel motion compensation with rounding (a+b+1)>>1.
30f15053 248 * this is an array[4][4] of motion compensation functions for 4
e5771f4f 249 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 250 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
251 * @param block destination where the result is stored
252 * @param pixels source
253 * @param line_size number of bytes in a horizontal line of block
254 * @param h height
255 */
669ac79c 256 op_pixels_func put_pixels_tab[4][4];
d518aebd
MN
257
258 /**
259 * Halfpel motion compensation with rounding (a+b+1)>>1.
115329f1 260 * This is an array[4][4] of motion compensation functions for 4
e5771f4f 261 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 262 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
263 * @param block destination into which the result is averaged (a+b+1)>>1
264 * @param pixels source
265 * @param line_size number of bytes in a horizontal line of block
266 * @param h height
267 */
da3b9756 268 op_pixels_func avg_pixels_tab[4][4];
d518aebd
MN
269
270 /**
271 * Halfpel motion compensation with no rounding (a+b)>>1.
30f15053 272 * this is an array[2][4] of motion compensation functions for 2
eb14c713 273 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 274 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
275 * @param block destination where the result is stored
276 * @param pixels source
277 * @param line_size number of bytes in a horizontal line of block
278 * @param h height
279 */
dbc56b39 280 op_pixels_func put_no_rnd_pixels_tab[4][4];
d518aebd
MN
281
282 /**
283 * Halfpel motion compensation with no rounding (a+b)>>1.
30f15053 284 * this is an array[2][4] of motion compensation functions for 2
eb14c713 285 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
5755c27f 286 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
d518aebd
MN
287 * @param block destination into which the result is averaged (a+b)>>1
288 * @param pixels source
289 * @param line_size number of bytes in a horizontal line of block
290 * @param h height
291 */
dbc56b39 292 op_pixels_func avg_no_rnd_pixels_tab[4][4];
115329f1 293
c0a0170c 294 void (*put_no_rnd_pixels_l2[2])(uint8_t *block/*align width (8 or 16)*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
115329f1 295
669ac79c
MN
296 /**
297 * Thirdpel motion compensation with rounding (a+b+1)>>1.
30f15053
LB
298 * this is an array[12] of motion compensation functions for the 9 thirdpe
299 * positions<br>
669ac79c
MN
300 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
301 * @param block destination where the result is stored
302 * @param pixels source
303 * @param line_size number of bytes in a horizontal line of block
304 * @param h height
305 */
306 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
da3b9756
MM
307 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
308
eb4b3dd3
ZK
309 qpel_mc_func put_qpel_pixels_tab[2][16];
310 qpel_mc_func avg_qpel_pixels_tab[2][16];
311 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
312 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
1457ab52 313 qpel_mc_func put_mspel_pixels_tab[8];
115329f1 314
0da71265 315 /**
30f15053 316 * h264 Chroma MC
0da71265
MN
317 */
318 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
319 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
c374691b
DC
320 /* This is really one func used in VC-1 decoding */
321 h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3];
8013da73 322 h264_chroma_mc_func avg_no_rnd_vc1_chroma_pixels_tab[3];
eb4b3dd3 323
80e44bc3
MN
324 qpel_mc_func put_h264_qpel_pixels_tab[4][16];
325 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
115329f1 326
2833fc46
LM
327 qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
328 qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
329
9f2d1b4f
LM
330 h264_weight_func weight_h264_pixels_tab[10];
331 h264_biweight_func biweight_h264_pixels_tab[10];
115329f1 332
b482e2d1
MN
333 /* AVS specific */
334 qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
335 qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
336 void (*cavs_filter_lv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
337 void (*cavs_filter_lh)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
338 void (*cavs_filter_cv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
339 void (*cavs_filter_ch)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
340 void (*cavs_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
341
bb198e19 342 me_cmp_func pix_abs[2][4];
115329f1 343
11f18faf 344 /* huffyuv specific */
11f18faf 345 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
4a9ca0a2 346 void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
1457ab52 347 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
84705403
MN
348 /**
349 * subtract huffyuv's variant of median prediction
350 * note, this might read from src1[-1], src2[-1]
351 */
352 void (*sub_hfyu_median_prediction)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w, int *left, int *left_top);
3daa434a 353 void (*add_hfyu_median_prediction)(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top);
4a9ca0a2
LM
354 /* this might write to dst[w] */
355 void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
96711ecf 356 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
42251a2a 357
2272e04a
LM
358 void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
359 void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
360 /* v/h_loop_filter_luma_intra: align 16 */
712ca84c
JGG
361 void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
362 void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
2272e04a
LM
363 void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
364 void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
365 void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
366 void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
3e20143e
LM
367 // h264_loop_filter_strength: simd only. the C version is inlined in h264.c
368 void (*h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
4f20b45f 369 int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);
115329f1 370
332f9ac4
MN
371 void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
372 void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
373
fdbbf2e0 374 void (*h261_loop_filter)(uint8_t *src, int stride);
c6148de2 375
9abc7e0f
MN
376 void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
377 void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
378
9971331d
DC
379 void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
380 void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
381
26f3ae82
AJ
382 void (*vp6_filter_diag4)(uint8_t *dst, uint8_t *src, int stride,
383 const int16_t *h_weights,const int16_t *v_weights);
384
eb4825b5 385 /* assume len is a multiple of 4, and arrays are 16-byte aligned */
2dac4acf 386 void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
ac2e5564 387 void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
6810b93a
LM
388 /* no alignment needed */
389 void (*flac_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
eb4825b5 390 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
06972056 391 void (*vector_fmul)(float *dst, const float *src, int len);
eb4825b5
LM
392 void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
393 /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
394 void (*vector_fmul_add_add)(float *dst, const float *src0, const float *src1, const float *src2, int src3, int len, int step);
f27e1d64
LM
395 /* assume len is a multiple of 4, and arrays are 16-byte aligned */
396 void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
911e21a3
LM
397 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
398 void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
50e23ae9 399 void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
eb4825b5
LM
400
401 /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
98145875 402 * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
560fa9bf 403 void (*float_to_int16)(int16_t *dst, const float *src, long len);
5eb0f2a4 404 void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);
2dac4acf 405
b0368839
MN
406 /* (I)DCT */
407 void (*fdct)(DCTELEM *block/* align 16*/);
10acc479 408 void (*fdct248)(DCTELEM *block/* align 16*/);
115329f1 409
4fb518c3
MN
410 /* IDCT really*/
411 void (*idct)(DCTELEM *block/* align 16*/);
115329f1 412
24641185 413 /**
77c92c2d 414 * block -> idct -> clip to unsigned 8 bit -> dest.
24641185 415 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
9846cbdb 416 * @param line_size size in bytes of a horizontal line of dest
24641185 417 */
b0368839 418 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
115329f1 419
24641185
MN
420 /**
421 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
9846cbdb 422 * @param line_size size in bytes of a horizontal line of dest
24641185 423 */
b0368839 424 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
115329f1 425
24641185 426 /**
77c92c2d 427 * idct input permutation.
05493021
MN
428 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
429 * IDCT)
430 * this permutation must be performed before the idct_put/add, note, normally this can be merged
431 * with the zigzag/alternate scan<br>
24641185
MN
432 * an example to avoid confusion:
433 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
434 * - (x -> referece dct -> reference idct -> x)
435 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
436 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
437 */
b0368839
MN
438 uint8_t idct_permutation[64];
439 int idct_permutation_type;
440#define FF_NO_IDCT_PERM 1
441#define FF_LIBMPEG2_IDCT_PERM 2
442#define FF_SIMPLE_IDCT_PERM 3
443#define FF_TRANSPOSE_IDCT_PERM 4
5773a746 444#define FF_PARTTRANS_IDCT_PERM 5
0e956ba2 445#define FF_SSE2_IDCT_PERM 6
b0368839 446
364a1797
MN
447 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
448 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
449#define BASIS_SHIFT 16
450#define RECON_SHIFT 6
115329f1 451
5a6a9e78 452 void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
cbcd6c8c 453#define EDGE_WIDTH 16
5a6a9e78 454
ea9f5d6f 455 /* h264 functions */
6a1846eb
MN
456 /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
457 NOTE!!! if you implement any of h264_idct_add, h264_idct_add16, h264_idct_add16intra, h264_idct_add8 then you must implement all of them
458 The reason for above, is that no 2 out of one list may use a different permutation.
459 */
7fadc32d
LM
460 void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
461 void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
462 void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
463 void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
f1862127 464 void (*h264_dct)(DCTELEM block[4][4]);
ac223859
MN
465 void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
466 void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
467 void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
468 void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
059715a4
RE
469
470 /* snow wavelet */
d593e329
MN
471 void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
472 void (*horizontal_compose97i)(IDWTELEM *b, int width);
9dd6c804 473 void (*inner_add_yblock)(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8);
513fbd8e
LM
474
475 void (*prefetch)(void *mem, int stride, int h);
54009d42
MN
476
477 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
64db55ae 478
bf4f19dc 479 /* mlp/truehd functions */
13bd2044
RP
480 void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
481 int firorder, int iirorder,
bf4f19dc
RP
482 unsigned int filter_shift, int32_t mask, int blocksize,
483 int32_t *sample_buffer);
484
64db55ae
KS
485 /* vc1 functions */
486 void (*vc1_inv_trans_8x8)(DCTELEM *b);
d2e45f33
KS
487 void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
488 void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
489 void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
4f717c69
JGG
490 void (*vc1_inv_trans_8x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
491 void (*vc1_inv_trans_8x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
492 void (*vc1_inv_trans_4x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
493 void (*vc1_inv_trans_4x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
61f5b14a
KS
494 void (*vc1_v_overlap)(uint8_t* src, int stride);
495 void (*vc1_h_overlap)(uint8_t* src, int stride);
3992526b
DC
496 void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq);
497 void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq);
498 void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq);
499 void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq);
500 void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq);
501 void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
64db55ae
KS
502 /* put 8x8 block with bicubic interpolation and quarterpel precision
503 * last argument is actually round value instead of height
504 */
505 op_pixels_func put_vc1_mspel_pixels_tab[16];
6cecd630 506 op_pixels_func avg_vc1_mspel_pixels_tab[16];
9abc7e0f
MN
507
508 /* intrax8 functions */
8a43317e
DB
509 void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
510 void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
9abc7e0f
MN
511 int * range, int * sum, int edges);
512
88c0536a
KS
513 /* ape functions */
514 /**
515 * Add contents of the second vector to the first one.
bb68f8a2 516 * @param len length of vectors, should be multiple of 16
88c0536a
KS
517 */
518 void (*add_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
519 /**
520 * Add contents of the second vector to the first one.
bb68f8a2 521 * @param len length of vectors, should be multiple of 16
88c0536a
KS
522 */
523 void (*sub_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
524 /**
525 * Calculate scalar product of two vectors.
bb68f8a2 526 * @param len length of vectors, should be multiple of 16
88c0536a
KS
527 * @param shift number of bits to discard from product
528 */
529 int32_t (*scalarproduct_int16)(int16_t *v1, int16_t *v2/*align 16*/, int len, int shift);
2d8a0815 530
11c0f9ec
KS
531 /* rv30 functions */
532 qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
533 qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
534
2d8a0815
KS
535 /* rv40 functions */
536 qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
537 qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
538 h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
539 h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
eb4b3dd3
ZK
540} DSPContext;
541
486497e0 542void dsputil_static_init(void);
b0368839 543void dsputil_init(DSPContext* p, AVCodecContext *avctx);
de6d9b64 544
6dc7d5da
MN
545int ff_check_alignment(void);
546
7801d21d
MN
547/**
548 * permute block according to permuatation.
549 * @param last last non zero element in scantable order
550 */
0c1a9eda 551void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
e0eac44e 552
622348f9
MN
553void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
554
bb270c08 555#define BYTE_VEC32(c) ((c)*0x01010101UL)
d8085ea7
MN
556
557static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
558{
559 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
560}
561
562static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
563{
564 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
565}
566
26efc54e
MN
567static inline int get_penalty_factor(int lambda, int lambda2, int type){
568 switch(type&0xFF){
569 default:
570 case FF_CMP_SAD:
571 return lambda>>FF_LAMBDA_SHIFT;
572 case FF_CMP_DCT:
573 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
574 case FF_CMP_W53:
575 return (4*lambda)>>(FF_LAMBDA_SHIFT);
576 case FF_CMP_W97:
577 return (2*lambda)>>(FF_LAMBDA_SHIFT);
578 case FF_CMP_SATD:
27c61ac5 579 case FF_CMP_DCT264:
26efc54e
MN
580 return (2*lambda)>>FF_LAMBDA_SHIFT;
581 case FF_CMP_RD:
582 case FF_CMP_PSNR:
583 case FF_CMP_SSE:
584 case FF_CMP_NSSE:
585 return lambda2>>FF_LAMBDA_SHIFT;
586 case FF_CMP_BIT:
587 return 1;
588 }
589}
590
24641185 591/**
77c92c2d 592 * Empty mmx state.
24641185
MN
593 * this must be called between any dsp function and float/double code.
594 * for example sin(); dsp->idct_put(); emms_c(); cos()
595 */
eb4b3dd3
ZK
596#define emms_c()
597
e629ab68
RD
598/* should be defined by architectures supporting
599 one or more MultiMedia extension */
600int mm_support(void);
e78516c6 601extern int mm_flags;
e629ab68 602
a1bee080 603void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
a2fc0f6a 604void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
a1bee080
MR
605void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
606void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
607void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
608void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
609void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
610void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
611void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
612
43c7c7c7 613#define DECLARE_ALIGNED_16(t, v) DECLARE_ALIGNED(16, t, v)
5b1b0147 614#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
92a69cf8 615
b250f9c6 616#if HAVE_MMX
de6d9b64 617
18f77016 618#undef emms_c
eb4b3dd3 619
de6d9b64
FB
620static inline void emms(void)
621{
be449fca 622 __asm__ volatile ("emms;":::"memory");
fb16b7e7
FB
623}
624
1457ab52 625
fb16b7e7
FB
626#define emms_c() \
627{\
82d1605f 628 if (mm_flags & FF_MM_MMX)\
fb16b7e7 629 emms();\
de6d9b64
FB
630}
631
b250f9c6 632#elif ARCH_ARM
3d03c0a2 633
b250f9c6 634#if HAVE_NEON
f1424cfd
MR
635# define STRIDE_ALIGN 16
636#endif
637
b250f9c6 638#elif ARCH_PPC
59925ef2 639
3237f731 640#define STRIDE_ALIGN 16
59925ef2 641
b250f9c6 642#elif HAVE_MMI
d46aba26 643
3237f731 644#define STRIDE_ALIGN 16
d46aba26 645
f67a10cd
AS
646#else
647
648#define mm_flags 0
649#define mm_support() 0
650
5c319d33 651#endif
de6d9b64 652
5c319d33
MR
653#ifndef STRIDE_ALIGN
654# define STRIDE_ALIGN 8
de6d9b64
FB
655#endif
656
43f1708f 657/* PSNR */
0c1a9eda 658void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
43f1708f
J
659 int orig_linesize[3], int coded_linesize,
660 AVCodecContext *avctx);
bb6f5690
FB
661
662/* FFT computation */
663
664/* NOTE: soon integer code will be added, so you must use the
665 FFTSample type */
666typedef float FFTSample;
667
668typedef struct FFTComplex {
669 FFTSample re, im;
670} FFTComplex;
671
672typedef struct FFTContext {
673 int nbits;
674 int inverse;
675 uint16_t *revtab;
676 FFTComplex *exptab;
677 FFTComplex *exptab1; /* only used by SSE code */
5d0ddd1a 678 FFTComplex *tmp_buf;
01b22147
MR
679 int mdct_size; /* size of MDCT (i.e. number of input data * 2) */
680 int mdct_bits; /* n = 2^nbits */
681 /* pre/post rotation tables */
682 FFTSample *tcos;
683 FFTSample *tsin;
5d0ddd1a 684 void (*fft_permute)(struct FFTContext *s, FFTComplex *z);
bb6f5690 685 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
01b22147
MR
686 void (*imdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
687 void (*imdct_half)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
688 void (*mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
f4863213 689 int split_radix;
94274b82
MR
690 int permutation;
691#define FF_MDCT_PERM_NONE 0
692#define FF_MDCT_PERM_INTERLEAVE 1
bb6f5690
FB
693} FFTContext;
694
faca5619 695extern FFTSample* const ff_cos_tabs[13];
68602540 696
8f05c995
AC
697/**
698 * Sets up a complex FFT.
699 * @param nbits log2 of the length of the input array
700 * @param inverse if 0 perform the forward transform, if 1 perform the inverse
701 */
68951ecf 702int ff_fft_init(FFTContext *s, int nbits, int inverse);
5d0ddd1a 703void ff_fft_permute_c(FFTContext *s, FFTComplex *z);
68951ecf 704void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
f4863213
MR
705
706void ff_fft_init_altivec(FFTContext *s);
707void ff_fft_init_mmx(FFTContext *s);
68336ea8 708void ff_fft_init_arm(FFTContext *s);
8d268a7d 709
8f05c995
AC
710/**
711 * Do the permutation needed BEFORE calling ff_fft_calc().
712 */
5d0ddd1a
LM
713static inline void ff_fft_permute(FFTContext *s, FFTComplex *z)
714{
715 s->fft_permute(s, z);
716}
8f05c995
AC
717/**
718 * Do a complex FFT with the parameters defined in ff_fft_init(). The
719 * input data must be permuted before. No 1.0/sqrt(n) normalization is done.
720 */
68951ecf 721static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)
bb6f5690
FB
722{
723 s->fft_calc(s, z);
724}
68951ecf 725void ff_fft_end(FFTContext *s);
bb6f5690
FB
726
727/* MDCT computation */
728
01b22147 729static inline void ff_imdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input)
d46ac5bf 730{
01b22147 731 s->imdct_calc(s, output, input);
d46ac5bf 732}
01b22147 733static inline void ff_imdct_half(FFTContext *s, FFTSample *output, const FFTSample *input)
d46ac5bf 734{
01b22147 735 s->imdct_half(s, output, input);
d46ac5bf
LM
736}
737
01b22147 738static inline void ff_mdct_calc(FFTContext *s, FFTSample *output,
46c32e26
MR
739 const FFTSample *input)
740{
01b22147 741 s->mdct_calc(s, output, input);
46c32e26
MR
742}
743
4eb7a735
RS
744/**
745 * Generate a Kaiser-Bessel Derived Window.
746 * @param window pointer to half window
3ed546fe
RS
747 * @param alpha determines window shape
748 * @param n size of half window
4eb7a735 749 */
3ed546fe 750void ff_kbd_window_init(float *window, float alpha, int n);
4eb7a735 751
9146e4d6
RS
752/**
753 * Generate a sine window.
754 * @param window pointer to half window
755 * @param n size of half window
756 */
757void ff_sine_window_init(float *window, int n);
6776061b
VS
758extern float ff_sine_32 [ 32];
759extern float ff_sine_64 [ 64];
69fc4da3
RS
760extern float ff_sine_128 [ 128];
761extern float ff_sine_256 [ 256];
762extern float ff_sine_512 [ 512];
763extern float ff_sine_1024[1024];
764extern float ff_sine_2048[2048];
336c2ca5 765extern float ff_sine_4096[4096];
6776061b 766extern float * const ff_sine_windows[13];
9146e4d6 767
01b22147
MR
768int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale);
769void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
770void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input);
771void ff_mdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
772void ff_mdct_end(FFTContext *s);
bb6f5690 773
68602540
AC
774/* Real Discrete Fourier Transform */
775
776enum RDFTransformType {
777 RDFT,
778 IRDFT,
779 RIDFT,
780 IRIDFT,
781};
782
783typedef struct {
784 int nbits;
785 int inverse;
786 int sign_convention;
787
788 /* pre/post rotation tables */
789 FFTSample *tcos;
790 FFTSample *tsin;
791 FFTContext fft;
792} RDFTContext;
793
794/**
795 * Sets up a real FFT.
796 * @param nbits log2 of the length of the input array
797 * @param trans the type of transform
798 */
799int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
800void ff_rdft_calc(RDFTContext *s, FFTSample *data);
801void ff_rdft_end(RDFTContext *s);
802
9fbd14ac 803#define WRAPPER8_16(name8, name16)\
bb198e19
MN
804static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
805 return name8(s, dst , src , stride, h)\
806 +name8(s, dst+8 , src+8 , stride, h);\
807}
808
9fbd14ac 809#define WRAPPER8_16_SQ(name8, name16)\
bb198e19
MN
810static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
811 int score=0;\
812 score +=name8(s, dst , src , stride, 8);\
813 score +=name8(s, dst+8 , src+8 , stride, 8);\
814 if(h==16){\
815 dst += 8*stride;\
816 src += 8*stride;\
817 score +=name8(s, dst , src , stride, 8);\
818 score +=name8(s, dst+8 , src+8 , stride, 8);\
819 }\
820 return score;\
1457ab52
MN
821}
822
49cef744 823
184fcc60 824static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
825{
826 int i;
827 for(i=0; i<h; i++)
828 {
905694d9 829 AV_WN16(dst , AV_RN16(src ));
49cef744
BF
830 dst+=dstStride;
831 src+=srcStride;
832 }
833}
834
184fcc60 835static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
836{
837 int i;
838 for(i=0; i<h; i++)
839 {
905694d9 840 AV_WN32(dst , AV_RN32(src ));
49cef744
BF
841 dst+=dstStride;
842 src+=srcStride;
843 }
844}
845
184fcc60 846static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
847{
848 int i;
849 for(i=0; i<h; i++)
850 {
905694d9
RS
851 AV_WN32(dst , AV_RN32(src ));
852 AV_WN32(dst+4 , AV_RN32(src+4 ));
49cef744
BF
853 dst+=dstStride;
854 src+=srcStride;
855 }
856}
857
184fcc60 858static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
859{
860 int i;
861 for(i=0; i<h; i++)
862 {
905694d9
RS
863 AV_WN32(dst , AV_RN32(src ));
864 AV_WN32(dst+4 , AV_RN32(src+4 ));
49cef744
BF
865 dst[8]= src[8];
866 dst+=dstStride;
867 src+=srcStride;
868 }
869}
870
184fcc60 871static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
872{
873 int i;
874 for(i=0; i<h; i++)
875 {
905694d9
RS
876 AV_WN32(dst , AV_RN32(src ));
877 AV_WN32(dst+4 , AV_RN32(src+4 ));
878 AV_WN32(dst+8 , AV_RN32(src+8 ));
879 AV_WN32(dst+12, AV_RN32(src+12));
49cef744
BF
880 dst+=dstStride;
881 src+=srcStride;
882 }
883}
884
184fcc60 885static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
49cef744
BF
886{
887 int i;
888 for(i=0; i<h; i++)
889 {
905694d9
RS
890 AV_WN32(dst , AV_RN32(src ));
891 AV_WN32(dst+4 , AV_RN32(src+4 ));
892 AV_WN32(dst+8 , AV_RN32(src+8 ));
893 AV_WN32(dst+12, AV_RN32(src+12));
49cef744
BF
894 dst[16]= src[16];
895 dst+=dstStride;
896 src+=srcStride;
897 }
898}
899
98790382 900#endif /* AVCODEC_DSPUTIL_H */