1/*
2 * H.264 IDCT
3 * Copyright (c) 2004-2011 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * H.264 IDCT.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28#include "bit_depth_template.c"
29
30#ifndef AVCODEC_H264IDCT_INTERNAL_H
31#define AVCODEC_H264IDCT_INTERNAL_H
32//FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
33static const uint8_t scan8[16*3]={
34 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
35 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
36 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
37 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
38 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
39 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
40 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
41 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
42 4+11*8, 5+11*8, 4+12*8, 5+12*8,
43 6+11*8, 7+11*8, 6+12*8, 7+12*8,
44 4+13*8, 5+13*8, 4+14*8, 5+14*8,
45 6+13*8, 7+13*8, 6+14*8, 7+14*8
46};
47#endif
48
49void FUNCC(ff_h264_idct_add)(uint8_t *_dst, DCTELEM *_block, int stride)
50{
51    int i;
52    pixel *dst = (pixel*)_dst;
53    dctcoef *block = (dctcoef*)_block;
54    stride /= sizeof(pixel);
55
56    block[0] += 1 << 5;
57
58    for(i=0; i<4; i++){
59        const int z0=  block[i + 4*0]     +  block[i + 4*2];
60        const int z1=  block[i + 4*0]     -  block[i + 4*2];
61        const int z2= (block[i + 4*1]>>1) -  block[i + 4*3];
62        const int z3=  block[i + 4*1]     + (block[i + 4*3]>>1);
63
64        block[i + 4*0]= z0 + z3;
65        block[i + 4*1]= z1 + z2;
66        block[i + 4*2]= z1 - z2;
67        block[i + 4*3]= z0 - z3;
68    }
69
70    for(i=0; i<4; i++){
71        const int z0=  block[0 + 4*i]     +  block[2 + 4*i];
72        const int z1=  block[0 + 4*i]     -  block[2 + 4*i];
73        const int z2= (block[1 + 4*i]>>1) -  block[3 + 4*i];
74        const int z3=  block[1 + 4*i]     + (block[3 + 4*i]>>1);
75
76        dst[i + 0*stride]= av_clip_pixel(dst[i + 0*stride] + ((z0 + z3) >> 6));
77        dst[i + 1*stride]= av_clip_pixel(dst[i + 1*stride] + ((z1 + z2) >> 6));
78        dst[i + 2*stride]= av_clip_pixel(dst[i + 2*stride] + ((z1 - z2) >> 6));
79        dst[i + 3*stride]= av_clip_pixel(dst[i + 3*stride] + ((z0 - z3) >> 6));
80    }
81}
82
83void FUNCC(ff_h264_idct8_add)(uint8_t *_dst, DCTELEM *_block, int stride){
84    int i;
85    pixel *dst = (pixel*)_dst;
86    dctcoef *block = (dctcoef*)_block;
87    stride /= sizeof(pixel);
88
89    block[0] += 32;
90
91    for( i = 0; i < 8; i++ )
92    {
93        const int a0 =  block[i+0*8] + block[i+4*8];
94        const int a2 =  block[i+0*8] - block[i+4*8];
95        const int a4 = (block[i+2*8]>>1) - block[i+6*8];
96        const int a6 = (block[i+6*8]>>1) + block[i+2*8];
97
98        const int b0 = a0 + a6;
99        const int b2 = a2 + a4;
100        const int b4 = a2 - a4;
101        const int b6 = a0 - a6;
102
103        const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
104        const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
105        const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
106        const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
107
108        const int b1 = (a7>>2) + a1;
109        const int b3 =  a3 + (a5>>2);
110        const int b5 = (a3>>2) - a5;
111        const int b7 =  a7 - (a1>>2);
112
113        block[i+0*8] = b0 + b7;
114        block[i+7*8] = b0 - b7;
115        block[i+1*8] = b2 + b5;
116        block[i+6*8] = b2 - b5;
117        block[i+2*8] = b4 + b3;
118        block[i+5*8] = b4 - b3;
119        block[i+3*8] = b6 + b1;
120        block[i+4*8] = b6 - b1;
121    }
122    for( i = 0; i < 8; i++ )
123    {
124        const int a0 =  block[0+i*8] + block[4+i*8];
125        const int a2 =  block[0+i*8] - block[4+i*8];
126        const int a4 = (block[2+i*8]>>1) - block[6+i*8];
127        const int a6 = (block[6+i*8]>>1) + block[2+i*8];
128
129        const int b0 = a0 + a6;
130        const int b2 = a2 + a4;
131        const int b4 = a2 - a4;
132        const int b6 = a0 - a6;
133
134        const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
135        const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
136        const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
137        const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
138
139        const int b1 = (a7>>2) + a1;
140        const int b3 =  a3 + (a5>>2);
141        const int b5 = (a3>>2) - a5;
142        const int b7 =  a7 - (a1>>2);
143
144        dst[i + 0*stride] = av_clip_pixel( dst[i + 0*stride] + ((b0 + b7) >> 6) );
145        dst[i + 1*stride] = av_clip_pixel( dst[i + 1*stride] + ((b2 + b5) >> 6) );
146        dst[i + 2*stride] = av_clip_pixel( dst[i + 2*stride] + ((b4 + b3) >> 6) );
147        dst[i + 3*stride] = av_clip_pixel( dst[i + 3*stride] + ((b6 + b1) >> 6) );
148        dst[i + 4*stride] = av_clip_pixel( dst[i + 4*stride] + ((b6 - b1) >> 6) );
149        dst[i + 5*stride] = av_clip_pixel( dst[i + 5*stride] + ((b4 - b3) >> 6) );
150        dst[i + 6*stride] = av_clip_pixel( dst[i + 6*stride] + ((b2 - b5) >> 6) );
151        dst[i + 7*stride] = av_clip_pixel( dst[i + 7*stride] + ((b0 - b7) >> 6) );
152    }
153}
154
155// assumes all AC coefs are 0
156void FUNCC(ff_h264_idct_dc_add)(uint8_t *_dst, DCTELEM *block, int stride){
157    int i, j;
158    int dc = (((dctcoef*)block)[0] + 32) >> 6;
159    pixel *dst = (pixel*)_dst;
160    stride /= sizeof(pixel);
161    for( j = 0; j < 4; j++ )
162    {
163        for( i = 0; i < 4; i++ )
164            dst[i] = av_clip_pixel( dst[i] + dc );
165        dst += stride;
166    }
167}
168
169void FUNCC(ff_h264_idct8_dc_add)(uint8_t *_dst, DCTELEM *block, int stride){
170    int i, j;
171    int dc = (((dctcoef*)block)[0] + 32) >> 6;
172    pixel *dst = (pixel*)_dst;
173    stride /= sizeof(pixel);
174    for( j = 0; j < 8; j++ )
175    {
176        for( i = 0; i < 8; i++ )
177            dst[i] = av_clip_pixel( dst[i] + dc );
178        dst += stride;
179    }
180}
181
182void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
183    int i;
184    for(i=0; i<16; i++){
185        int nnz = nnzc[ scan8[i] ];
186        if(nnz){
187            if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
188            else                                  FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
189        }
190    }
191}
192
193void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
194    int i;
195    for(i=0; i<16; i++){
196        if(nnzc[ scan8[i] ])             FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
197        else if(((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
198    }
199}
200
201void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
202    int i;
203    for(i=0; i<16; i+=4){
204        int nnz = nnzc[ scan8[i] ];
205        if(nnz){
206            if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct8_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
207            else                                  FUNCC(ff_h264_idct8_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
208        }
209    }
210}
211
212void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
213    int i, j;
214    for(j=1; j<3; j++){
215        for(i=j*16; i<j*16+4; i++){
216            if(nnzc[ scan8[i] ])
217                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
218            else if(((dctcoef*)block)[i*16])
219                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
220        }
221    }
222}
223
224void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[15*8]){
225    int i, j;
226
227    for(j=1; j<3; j++){
228        for(i=j*16; i<j*16+4; i++){
229            if(nnzc[ scan8[i] ])
230                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
231            else if(((dctcoef*)block)[i*16])
232                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
233        }
234    }
235
236    for(j=1; j<3; j++){
237        for(i=j*16+4; i<j*16+8; i++){
238            if(nnzc[ scan8[i+4] ])
239                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
240            else if(((dctcoef*)block)[i*16])
241                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
242        }
243    }
244}
245
246/**
247 * IDCT transforms the 16 dc values and dequantizes them.
248 * @param qmul quantization parameter
249 */
250void FUNCC(ff_h264_luma_dc_dequant_idct)(DCTELEM *_output, DCTELEM *_input, int qmul){
251#define stride 16
252    int i;
253    int temp[16];
254    static const uint8_t x_offset[4]={0, 2*stride, 8*stride, 10*stride};
255    dctcoef *input = (dctcoef*)_input;
256    dctcoef *output = (dctcoef*)_output;
257
258    for(i=0; i<4; i++){
259        const int z0= input[4*i+0] + input[4*i+1];
260        const int z1= input[4*i+0] - input[4*i+1];
261        const int z2= input[4*i+2] - input[4*i+3];
262        const int z3= input[4*i+2] + input[4*i+3];
263
264        temp[4*i+0]= z0+z3;
265        temp[4*i+1]= z0-z3;
266        temp[4*i+2]= z1-z2;
267        temp[4*i+3]= z1+z2;
268    }
269
270    for(i=0; i<4; i++){
271        const int offset= x_offset[i];
272        const int z0= temp[4*0+i] + temp[4*2+i];
273        const int z1= temp[4*0+i] - temp[4*2+i];
274        const int z2= temp[4*1+i] - temp[4*3+i];
275        const int z3= temp[4*1+i] + temp[4*3+i];
276
277        output[stride* 0+offset]= ((((z0 + z3)*qmul + 128 ) >> 8));
278        output[stride* 1+offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
279        output[stride* 4+offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
280        output[stride* 5+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
281    }
282#undef stride
283}
284
285void FUNCC(ff_h264_chroma422_dc_dequant_idct)(DCTELEM *_block, int qmul){
286    const int stride= 16*2;
287    const int xStride= 16;
288    int i;
289    int temp[8];
290    static const uint8_t x_offset[2]={0, 16};
291    dctcoef *block = (dctcoef*)_block;
292
293    for(i=0; i<4; i++){
294        temp[2*i+0] = block[stride*i + xStride*0] + block[stride*i + xStride*1];
295        temp[2*i+1] = block[stride*i + xStride*0] - block[stride*i + xStride*1];
296    }
297
298    for(i=0; i<2; i++){
299        const int offset= x_offset[i];
300        const int z0= temp[2*0+i] + temp[2*2+i];
301        const int z1= temp[2*0+i] - temp[2*2+i];
302        const int z2= temp[2*1+i] - temp[2*3+i];
303        const int z3= temp[2*1+i] + temp[2*3+i];
304
305        block[stride*0+offset]= ((z0 + z3)*qmul + 128) >> 8;
306        block[stride*1+offset]= ((z1 + z2)*qmul + 128) >> 8;
307        block[stride*2+offset]= ((z1 - z2)*qmul + 128) >> 8;
308        block[stride*3+offset]= ((z0 - z3)*qmul + 128) >> 8;
309    }
310}
311
312void FUNCC(ff_h264_chroma_dc_dequant_idct)(DCTELEM *_block, int qmul){
313    const int stride= 16*2;
314    const int xStride= 16;
315    int a,b,c,d,e;
316    dctcoef *block = (dctcoef*)_block;
317
318    a= block[stride*0 + xStride*0];
319    b= block[stride*0 + xStride*1];
320    c= block[stride*1 + xStride*0];
321    d= block[stride*1 + xStride*1];
322
323    e= a-b;
324    a= a+b;
325    b= c-d;
326    c= c+d;
327
328    block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
329    block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
330    block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
331    block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
332}
333