summaryrefslogtreecommitdiff
path: root/apps/codecs/libcook/dsputil.h
diff options
context:
space:
mode:
Diffstat (limited to 'apps/codecs/libcook/dsputil.h')
-rw-r--r--apps/codecs/libcook/dsputil.h908
1 files changed, 0 insertions, 908 deletions
diff --git a/apps/codecs/libcook/dsputil.h b/apps/codecs/libcook/dsputil.h
index 4573c17a62..e69de29bb2 100644
--- a/apps/codecs/libcook/dsputil.h
+++ b/apps/codecs/libcook/dsputil.h
@@ -1,908 +0,0 @@
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 FFmpeg.
7 *
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file libavcodec/dsputil.h
25 * DSP utils.
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
28 */
29
30#ifndef AVCODEC_DSPUTIL_H
31#define AVCODEC_DSPUTIL_H
32
33#include "libavutil/intreadwrite.h"
34#include "avcodec.h"
35
36
37//#define DEBUG
38/* dct code */
39#if 0 /*MT : DELETE THIS LINE.*/
40typedef short DCTELEM;
41typedef int DWTELEM;
42typedef short IDWTELEM;
43
44void fdct_ifast (DCTELEM *data);
45void fdct_ifast248 (DCTELEM *data);
46void ff_jpeg_fdct_islow (DCTELEM *data);
47void ff_fdct248_islow (DCTELEM *data);
48
49void j_rev_dct (DCTELEM *data);
50void j_rev_dct4 (DCTELEM *data);
51void j_rev_dct2 (DCTELEM *data);
52void j_rev_dct1 (DCTELEM *data);
53void ff_wmv2_idct_c(DCTELEM *data);
54
55void ff_fdct_mmx(DCTELEM *block);
56void ff_fdct_mmx2(DCTELEM *block);
57void ff_fdct_sse2(DCTELEM *block);
58
59void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
60void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride);
61void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
62void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
63void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
64void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
65void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
66void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
67void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
68void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
69
70void ff_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
71 const float *src2, int src3, int blocksize, int step);
72void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
73 const float *win, float add_bias, int len);
74void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
75void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels);
76
77/* encoding scans */
78extern const uint8_t ff_alternate_horizontal_scan[64];
79extern const uint8_t ff_alternate_vertical_scan[64];
80extern const uint8_t ff_zigzag_direct[64];
81extern const uint8_t ff_zigzag248_direct[64];
82
83/* pixel operations */
84#define MAX_NEG_CROP 1024
85
86/* temporary */
87extern uint32_t ff_squareTbl[512];
88extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
89
90/* VP3 DSP functions */
91void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
92void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
93void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
94
95void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
96void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
97
98/* VP6 DSP functions */
99void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
100 const int16_t *h_weights, const int16_t *v_weights);
101
102/* 1/2^n downscaling functions from imgconvert.c */
103void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
104void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
105void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
106void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
107
108void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
109 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
110
111/* minimum alignment rules ;)
112If you notice errors in the align stuff, need more alignment for some ASM code
113for some CPU or need to use a function with less aligned data then send a mail
114to the ffmpeg-devel mailing list, ...
115
116!warning These alignments might not match reality, (missing attribute((align))
117stuff somewhere possible).
118I (Michael) did not check them, these are just the alignments which I think
119could be reached easily ...
120
121!future video codecs might need functions with less strict alignment
122*/
123
124/*
125void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
126void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
127void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
128void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
129void clear_blocks_c(DCTELEM *blocks);
130*/
131
132/* add and put pixel (decoding) */
133// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
134//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
135typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
136typedef 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);
137typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
138typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
139typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
140typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
141
142#define DEF_OLD_QPEL(name)\
143void ff_put_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
144void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
145void ff_avg_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
146
147DEF_OLD_QPEL(qpel16_mc11_old_c)
148DEF_OLD_QPEL(qpel16_mc31_old_c)
149DEF_OLD_QPEL(qpel16_mc12_old_c)
150DEF_OLD_QPEL(qpel16_mc32_old_c)
151DEF_OLD_QPEL(qpel16_mc13_old_c)
152DEF_OLD_QPEL(qpel16_mc33_old_c)
153DEF_OLD_QPEL(qpel8_mc11_old_c)
154DEF_OLD_QPEL(qpel8_mc31_old_c)
155DEF_OLD_QPEL(qpel8_mc12_old_c)
156DEF_OLD_QPEL(qpel8_mc32_old_c)
157DEF_OLD_QPEL(qpel8_mc13_old_c)
158DEF_OLD_QPEL(qpel8_mc33_old_c)
159
160#define CALL_2X_PIXELS(a, b, n)\
161static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
162 b(block , pixels , line_size, h);\
163 b(block+n, pixels+n, line_size, h);\
164}
165
166/* motion estimation */
167// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
168// although currently h<4 is not used as functions with width <8 are neither used nor implemented
169typedef 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))*/;
170
171
172// for snow slices
173typedef struct slice_buffer_s slice_buffer;
174
175/**
176 * Scantable.
177 */
178typedef struct ScanTable{
179 const uint8_t *scantable;
180 uint8_t permutated[64];
181 uint8_t raster_end[64];
182#if ARCH_PPC
183 /** Used by dct_quantize_altivec to find last-non-zero */
184 DECLARE_ALIGNED(16, uint8_t, inverse[64]);
185#endif
186} ScanTable;
187
188void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
189
190void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize,
191 int block_w, int block_h,
192 int src_x, int src_y, int w, int h);
193
194/**
195 * DSPContext.
196 */
197typedef struct DSPContext {
198 /* pixel ops : interface with DCT */
199 void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
200 void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
201 void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
202 void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
203 void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
204 void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
205 void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
206 int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
207 /**
208 * translational global motion compensation.
209 */
210 void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
211 /**
212 * global motion compensation.
213 */
214 void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
215 int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
216 void (*clear_block)(DCTELEM *block/*align 16*/);
217 void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
218 int (*pix_sum)(uint8_t * pix, int line_size);
219 int (*pix_norm1)(uint8_t * pix, int line_size);
220// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
221
222 me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
223 me_cmp_func sse[6];
224 me_cmp_func hadamard8_diff[6];
225 me_cmp_func dct_sad[6];
226 me_cmp_func quant_psnr[6];
227 me_cmp_func bit[6];
228 me_cmp_func rd[6];
229 me_cmp_func vsad[6];
230 me_cmp_func vsse[6];
231 me_cmp_func nsse[6];
232 me_cmp_func w53[6];
233 me_cmp_func w97[6];
234 me_cmp_func dct_max[6];
235 me_cmp_func dct264_sad[6];
236
237 me_cmp_func me_pre_cmp[6];
238 me_cmp_func me_cmp[6];
239 me_cmp_func me_sub_cmp[6];
240 me_cmp_func mb_cmp[6];
241 me_cmp_func ildct_cmp[6]; //only width 16 used
242 me_cmp_func frame_skip_cmp[6]; //only width 8 used
243
244 int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
245 int size);
246
247 /**
248 * Halfpel motion compensation with rounding (a+b+1)>>1.
249 * this is an array[4][4] of motion compensation functions for 4
250 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
251 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
252 * @param block destination where the result is stored
253 * @param pixels source
254 * @param line_size number of bytes in a horizontal line of block
255 * @param h height
256 */
257 op_pixels_func put_pixels_tab[4][4];
258
259 /**
260 * Halfpel motion compensation with rounding (a+b+1)>>1.
261 * This is an array[4][4] of motion compensation functions for 4
262 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
263 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
264 * @param block destination into which the result is averaged (a+b+1)>>1
265 * @param pixels source
266 * @param line_size number of bytes in a horizontal line of block
267 * @param h height
268 */
269 op_pixels_func avg_pixels_tab[4][4];
270
271 /**
272 * Halfpel motion compensation with no rounding (a+b)>>1.
273 * this is an array[2][4] of motion compensation functions for 2
274 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
275 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
276 * @param block destination where the result is stored
277 * @param pixels source
278 * @param line_size number of bytes in a horizontal line of block
279 * @param h height
280 */
281 op_pixels_func put_no_rnd_pixels_tab[4][4];
282
283 /**
284 * Halfpel motion compensation with no rounding (a+b)>>1.
285 * this is an array[2][4] of motion compensation functions for 2
286 * horizontal blocksizes (8,16) and the 4 halfpel positions<br>
287 * *pixels_tab[ 0->16xH 1->8xH ][ xhalfpel + 2*yhalfpel ]
288 * @param block destination into which the result is averaged (a+b)>>1
289 * @param pixels source
290 * @param line_size number of bytes in a horizontal line of block
291 * @param h height
292 */
293 op_pixels_func avg_no_rnd_pixels_tab[4][4];
294
295 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);
296
297 /**
298 * Thirdpel motion compensation with rounding (a+b+1)>>1.
299 * this is an array[12] of motion compensation functions for the 9 thirdpe
300 * positions<br>
301 * *pixels_tab[ xthirdpel + 4*ythirdpel ]
302 * @param block destination where the result is stored
303 * @param pixels source
304 * @param line_size number of bytes in a horizontal line of block
305 * @param h height
306 */
307 tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
308 tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
309
310 qpel_mc_func put_qpel_pixels_tab[2][16];
311 qpel_mc_func avg_qpel_pixels_tab[2][16];
312 qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
313 qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
314 qpel_mc_func put_mspel_pixels_tab[8];
315
316 /**
317 * h264 Chroma MC
318 */
319 h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
320 /* This is really one func used in VC-1 decoding */
321 h264_chroma_mc_func put_no_rnd_h264_chroma_pixels_tab[3];
322 h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
323
324 qpel_mc_func put_h264_qpel_pixels_tab[4][16];
325 qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
326
327 qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
328 qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
329
330 h264_weight_func weight_h264_pixels_tab[10];
331 h264_biweight_func biweight_h264_pixels_tab[10];
332
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
342 me_cmp_func pix_abs[2][4];
343
344 /* huffyuv specific */
345 void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
346 void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
347 void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
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);
353 void (*add_hfyu_median_prediction)(uint8_t *dst, uint8_t *top, uint8_t *diff, int w, int *left, int *left_top);
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);
356 void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
357
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 */
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);
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);
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],
369 int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);
370
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
374 void (*h261_loop_filter)(uint8_t *src, int stride);
375
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
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
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
385 /* assume len is a multiple of 4, and arrays are 16-byte aligned */
386 void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
387 void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
388 /* no alignment needed */
389 void (*flac_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
390 /* assume len is a multiple of 8, and arrays are 16-byte aligned */
391 void (*vector_fmul)(float *dst, const float *src, int len);
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);
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);
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);
399
400 /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
401 * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
402 void (*float_to_int16)(int16_t *dst, const float *src, long len);
403 void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);
404
405 /* (I)DCT */
406 void (*fdct)(DCTELEM *block/* align 16*/);
407 void (*fdct248)(DCTELEM *block/* align 16*/);
408
409 /* IDCT really*/
410 void (*idct)(DCTELEM *block/* align 16*/);
411
412 /**
413 * block -> idct -> clip to unsigned 8 bit -> dest.
414 * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
415 * @param line_size size in bytes of a horizontal line of dest
416 */
417 void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
418
419 /**
420 * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
421 * @param line_size size in bytes of a horizontal line of dest
422 */
423 void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
424
425 /**
426 * idct input permutation.
427 * several optimized IDCTs need a permutated input (relative to the normal order of the reference
428 * IDCT)
429 * this permutation must be performed before the idct_put/add, note, normally this can be merged
430 * with the zigzag/alternate scan<br>
431 * an example to avoid confusion:
432 * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
433 * - (x -> referece dct -> reference idct -> x)
434 * - (x -> referece dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
435 * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
436 */
437 uint8_t idct_permutation[64];
438 int idct_permutation_type;
439#define FF_NO_IDCT_PERM 1
440#define FF_LIBMPEG2_IDCT_PERM 2
441#define FF_SIMPLE_IDCT_PERM 3
442#define FF_TRANSPOSE_IDCT_PERM 4
443#define FF_PARTTRANS_IDCT_PERM 5
444#define FF_SSE2_IDCT_PERM 6
445
446 int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
447 void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
448#define BASIS_SHIFT 16
449#define RECON_SHIFT 6
450
451 void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
452#define EDGE_WIDTH 16
453
454 /* h264 functions */
455 /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
456 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
457 The reason for above, is that no 2 out of one list may use a different permutation.
458 */
459 void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
460 void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
461 void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
462 void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
463 void (*h264_dct)(DCTELEM block[4][4]);
464 void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
465 void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
466 void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
467 void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
468
469 /* snow wavelet */
470 void (*vertical_compose97i)(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width);
471 void (*horizontal_compose97i)(IDWTELEM *b, int width);
472 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);
473
474 void (*prefetch)(void *mem, int stride, int h);
475
476 void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
477
478 /* vc1 functions */
479 void (*vc1_inv_trans_8x8)(DCTELEM *b);
480 void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
481 void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
482 void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
483 void (*vc1_v_overlap)(uint8_t* src, int stride);
484 void (*vc1_h_overlap)(uint8_t* src, int stride);
485 /* put 8x8 block with bicubic interpolation and quarterpel precision
486 * last argument is actually round value instead of height
487 */
488 op_pixels_func put_vc1_mspel_pixels_tab[16];
489
490 /* intrax8 functions */
491 void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
492 void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
493 int * range, int * sum, int edges);
494
495 /* ape functions */
496 /**
497 * Add contents of the second vector to the first one.
498 * @param len length of vectors, should be multiple of 16
499 */
500 void (*add_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
501 /**
502 * Add contents of the second vector to the first one.
503 * @param len length of vectors, should be multiple of 16
504 */
505 void (*sub_int16)(int16_t *v1/*align 16*/, int16_t *v2, int len);
506 /**
507 * Calculate scalar product of two vectors.
508 * @param len length of vectors, should be multiple of 16
509 * @param shift number of bits to discard from product
510 */
511 int32_t (*scalarproduct_int16)(int16_t *v1, int16_t *v2/*align 16*/, int len, int shift);
512
513 /* rv30 functions */
514 qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
515 qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
516
517 /* rv40 functions */
518 qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
519 qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
520 h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
521 h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
522} DSPContext;
523
524void dsputil_static_init(void);
525void dsputil_init(DSPContext* p, AVCodecContext *avctx);
526
527int ff_check_alignment(void);
528
529/**
530 * permute block according to permuatation.
531 * @param last last non zero element in scantable order
532 */
533void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
534
535void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
536
537#define BYTE_VEC32(c) ((c)*0x01010101UL)
538
539static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
540{
541 return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
542}
543
544static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
545{
546 return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
547}
548
549static inline int get_penalty_factor(int lambda, int lambda2, int type){
550 switch(type&0xFF){
551 default:
552 case FF_CMP_SAD:
553 return lambda>>FF_LAMBDA_SHIFT;
554 case FF_CMP_DCT:
555 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
556 case FF_CMP_W53:
557 return (4*lambda)>>(FF_LAMBDA_SHIFT);
558 case FF_CMP_W97:
559 return (2*lambda)>>(FF_LAMBDA_SHIFT);
560 case FF_CMP_SATD:
561 case FF_CMP_DCT264:
562 return (2*lambda)>>FF_LAMBDA_SHIFT;
563 case FF_CMP_RD:
564 case FF_CMP_PSNR:
565 case FF_CMP_SSE:
566 case FF_CMP_NSSE:
567 return lambda2>>FF_LAMBDA_SHIFT;
568 case FF_CMP_BIT:
569 return 1;
570 }
571}
572
573/**
574 * Empty mmx state.
575 * this must be called between any dsp function and float/double code.
576 * for example sin(); dsp->idct_put(); emms_c(); cos()
577 */
578#define emms_c()
579
580/* should be defined by architectures supporting
581 one or more MultiMedia extension */
582int mm_support(void);
583
584void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
585void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
586void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
587void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
588void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
589void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
590void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
591void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
592void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
593
594#endif /*MT : DELETE THIS LINE ONLY. */
595#define DECLARE_ALIGNED_16(t, v) DECLARE_ALIGNED(16, t, v)
596
597#if 0 /*MT : DELETE THIS LINE ONLY. */
598#if HAVE_MMX
599
600#undef emms_c
601
602extern int mm_flags;
603
604void add_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
605void put_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
606void put_signed_pixels_clamped_mmx(const DCTELEM *block, uint8_t *pixels, int line_size);
607
608static inline void emms(void)
609{
610 __asm__ volatile ("emms;":::"memory");
611}
612
613
614#define emms_c() \
615{\
616 if (mm_flags & FF_MM_MMX)\
617 emms();\
618}
619
620void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx);
621
622#elif ARCH_ARM
623
624extern int mm_flags;
625
626#if HAVE_NEON
627# define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
628# define STRIDE_ALIGN 16
629#endif
630
631#elif ARCH_PPC
632
633extern int mm_flags;
634
635#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
636#define STRIDE_ALIGN 16
637
638#elif HAVE_MMI
639
640#define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(16, t, v)
641#define STRIDE_ALIGN 16
642
643#else
644
645#define mm_flags 0
646#define mm_support() 0
647
648#endif
649
650#endif /* MT : DELETE THIS LINE ONLY */
651#ifndef DECLARE_ALIGNED_8
652# define DECLARE_ALIGNED_8(t, v) DECLARE_ALIGNED(8, t, v)
653#endif
654
655#if 0 /* MT : DELETE THIS LINE ONLY */
656#ifndef STRIDE_ALIGN
657# define STRIDE_ALIGN 8
658#endif
659
660/* PSNR */
661void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
662 int orig_linesize[3], int coded_linesize,
663 AVCodecContext *avctx);
664
665#endif /*MT : DELETE THIS LINE.*/
666/* FFT computation */
667
668/* NOTE: soon integer code will be added, so you must use the
669 FFTSample type */
670typedef float FFTSample;
671
672struct MDCTContext;
673
674typedef struct FFTComplex {
675 FFTSample re, im;
676} FFTComplex;
677
678typedef struct FFTContext {
679 int nbits;
680 int inverse;
681 uint16_t *revtab;
682 FFTComplex *exptab;
683 FFTComplex *exptab1; /* only used by SSE code */
684 FFTComplex *tmp_buf;
685 void (*fft_permute)(struct FFTContext *s, FFTComplex *z);
686 void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
687 void (*imdct_calc)(struct MDCTContext *s, FFTSample *output, const FFTSample *input);
688 void (*imdct_half)(struct MDCTContext *s, FFTSample *output, const FFTSample *input);
689} FFTContext;
690
691extern FFTSample* ff_cos_tabs[13];
692
693/**
694 * Sets up a complex FFT.
695 * @param nbits log2 of the length of the input array
696 * @param inverse if 0 perform the forward transform, if 1 perform the inverse
697 */
698int ff_fft_init(FFTContext *s, int nbits, int inverse);
699void ff_fft_permute_c(FFTContext *s, FFTComplex *z);
700void ff_fft_permute_sse(FFTContext *s, FFTComplex *z);
701void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
702void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
703void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
704void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z);
705void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z);
706
707/**
708 * Do the permutation needed BEFORE calling ff_fft_calc().
709 */
710static inline void ff_fft_permute(FFTContext *s, FFTComplex *z)
711{
712 s->fft_permute(s, z);
713}
714/**
715 * Do a complex FFT with the parameters defined in ff_fft_init(). The
716 * input data must be permuted before. No 1.0/sqrt(n) normalization is done.
717 */
718static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)
719{
720 s->fft_calc(s, z);
721}
722void ff_fft_end(FFTContext *s);
723
724#endif /*MT : DELETE THIS LINE.*/
725/* MDCT computation */
726
727typedef struct MDCTContext {
728 int n; /* size of MDCT (i.e. number of input data * 2) */
729 int nbits; /* n = 2^nbits */
730 /* pre/post rotation tables */
731 FFTSample *tcos;
732 FFTSample *tsin;
733 FFTContext fft;
734} MDCTContext;
735
736static inline void ff_imdct_calc(MDCTContext *s, FFTSample *output, const FFTSample *input)
737{
738 s->fft.imdct_calc(s, output, input);
739}
740static inline void ff_imdct_half(MDCTContext *s, FFTSample *output, const FFTSample *input)
741{
742 s->fft.imdct_half(s, output, input);
743}
744
745#if 0 /* MT : DELETE THIS LINE. */
746/**
747 * Generate a Kaiser-Bessel Derived Window.
748 * @param window pointer to half window
749 * @param alpha determines window shape
750 * @param n size of half window
751 */
752void ff_kbd_window_init(float *window, float alpha, int n);
753#endif /* MT : DELETE THIS LINE.*/
754
755/**
756 * Generate a sine window.
757 * @param window pointer to half window
758 * @param n size of half window
759 */
760void ff_sine_window_init(float *window, int n);
761extern float ff_sine_128 [ 128];
762extern float ff_sine_256 [ 256];
763extern float ff_sine_512 [ 512];
764extern float ff_sine_1024[1024];
765extern float ff_sine_2048[2048];
766extern float ff_sine_4096[4096];
767extern float *ff_sine_windows[6];
768
769int ff_mdct_init(MDCTContext *s, int nbits, int inverse);
770void ff_imdct_calc_c(MDCTContext *s, FFTSample *output, const FFTSample *input);
771void ff_imdct_half_c(MDCTContext *s, FFTSample *output, const FFTSample *input);
772void ff_imdct_calc_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input);
773void ff_imdct_half_3dn(MDCTContext *s, FFTSample *output, const FFTSample *input);
774void ff_imdct_calc_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input);
775void ff_imdct_half_3dn2(MDCTContext *s, FFTSample *output, const FFTSample *input);
776void ff_imdct_calc_sse(MDCTContext *s, FFTSample *output, const FFTSample *input);
777void ff_imdct_half_sse(MDCTContext *s, FFTSample *output, const FFTSample *input);
778void ff_mdct_calc(MDCTContext *s, FFTSample *out, const FFTSample *input);
779void ff_mdct_end(MDCTContext *s);
780
781#if 0 /* MT : DELETE THIS LINE.*/
782/* Real Discrete Fourier Transform */
783
784enum RDFTransformType {
785 RDFT,
786 IRDFT,
787 RIDFT,
788 IRIDFT,
789};
790
791typedef struct {
792 int nbits;
793 int inverse;
794 int sign_convention;
795
796 /* pre/post rotation tables */
797 FFTSample *tcos;
798 FFTSample *tsin;
799 FFTContext fft;
800} RDFTContext;
801
802/**
803 * Sets up a real FFT.
804 * @param nbits log2 of the length of the input array
805 * @param trans the type of transform
806 */
807int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
808void ff_rdft_calc(RDFTContext *s, FFTSample *data);
809void ff_rdft_end(RDFTContext *s);
810
811#define WRAPPER8_16(name8, name16)\
812static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
813 return name8(s, dst , src , stride, h)\
814 +name8(s, dst+8 , src+8 , stride, h);\
815}
816
817#define WRAPPER8_16_SQ(name8, name16)\
818static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
819 int score=0;\
820 score +=name8(s, dst , src , stride, 8);\
821 score +=name8(s, dst+8 , src+8 , stride, 8);\
822 if(h==16){\
823 dst += 8*stride;\
824 src += 8*stride;\
825 score +=name8(s, dst , src , stride, 8);\
826 score +=name8(s, dst+8 , src+8 , stride, 8);\
827 }\
828 return score;\
829}
830
831
832static inline void copy_block2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
833{
834 int i;
835 for(i=0; i<h; i++)
836 {
837 AV_WN16(dst , AV_RN16(src ));
838 dst+=dstStride;
839 src+=srcStride;
840 }
841}
842
843static inline void copy_block4(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
844{
845 int i;
846 for(i=0; i<h; i++)
847 {
848 AV_WN32(dst , AV_RN32(src ));
849 dst+=dstStride;
850 src+=srcStride;
851 }
852}
853
854static inline void copy_block8(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
855{
856 int i;
857 for(i=0; i<h; i++)
858 {
859 AV_WN32(dst , AV_RN32(src ));
860 AV_WN32(dst+4 , AV_RN32(src+4 ));
861 dst+=dstStride;
862 src+=srcStride;
863 }
864}
865
866static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
867{
868 int i;
869 for(i=0; i<h; i++)
870 {
871 AV_WN32(dst , AV_RN32(src ));
872 AV_WN32(dst+4 , AV_RN32(src+4 ));
873 dst[8]= src[8];
874 dst+=dstStride;
875 src+=srcStride;
876 }
877}
878
879static inline void copy_block16(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
880{
881 int i;
882 for(i=0; i<h; i++)
883 {
884 AV_WN32(dst , AV_RN32(src ));
885 AV_WN32(dst+4 , AV_RN32(src+4 ));
886 AV_WN32(dst+8 , AV_RN32(src+8 ));
887 AV_WN32(dst+12, AV_RN32(src+12));
888 dst+=dstStride;
889 src+=srcStride;
890 }
891}
892
893static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)
894{
895 int i;
896 for(i=0; i<h; i++)
897 {
898 AV_WN32(dst , AV_RN32(src ));
899 AV_WN32(dst+4 , AV_RN32(src+4 ));
900 AV_WN32(dst+8 , AV_RN32(src+8 ));
901 AV_WN32(dst+12, AV_RN32(src+12));
902 dst[16]= src[16];
903 dst+=dstStride;
904 src+=srcStride;
905 }
906}
907
908#endif /* AVCODEC_DSPUTIL_H */
generated by cgit v1.2.3 (git 2.39.1) at 2024-09-24 12:30:07 +0000