1/*
2 * H.264 IDCT
3 * Copyright (c) 2004-2011 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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#include "libavutil/common.h"
30#include "h264.h"
31#include "h264idct.h"
32
33void FUNCC(ff_h264_idct_add)(uint8_t *_dst, int16_t *_block, int stride)
34{
35    int i;
36    pixel *dst = (pixel*)_dst;
37    dctcoef *block = (dctcoef*)_block;
38    stride >>= sizeof(pixel)-1;
39
40    block[0] += 1 << 5;
41
42    for(i=0; i<4; i++){
43        const int z0=  block[i + 4*0]     +  block[i + 4*2];
44        const int z1=  block[i + 4*0]     -  block[i + 4*2];
45        const int z2= (block[i + 4*1]>>1) -  block[i + 4*3];
46        const int z3=  block[i + 4*1]     + (block[i + 4*3]>>1);
47
48        block[i + 4*0]= z0 + z3;
49        block[i + 4*1]= z1 + z2;
50        block[i + 4*2]= z1 - z2;
51        block[i + 4*3]= z0 - z3;
52    }
53
54    for(i=0; i<4; i++){
55        const int z0=  block[0 + 4*i]     +  block[2 + 4*i];
56        const int z1=  block[0 + 4*i]     -  block[2 + 4*i];
57        const int z2= (block[1 + 4*i]>>1) -  block[3 + 4*i];
58        const int z3=  block[1 + 4*i]     + (block[3 + 4*i]>>1);
59
60        dst[i + 0*stride]= av_clip_pixel(dst[i + 0*stride] + ((z0 + z3) >> 6));
61        dst[i + 1*stride]= av_clip_pixel(dst[i + 1*stride] + ((z1 + z2) >> 6));
62        dst[i + 2*stride]= av_clip_pixel(dst[i + 2*stride] + ((z1 - z2) >> 6));
63        dst[i + 3*stride]= av_clip_pixel(dst[i + 3*stride] + ((z0 - z3) >> 6));
64    }
65
66    memset(block, 0, 16 * sizeof(dctcoef));
67}
68
69void FUNCC(ff_h264_idct8_add)(uint8_t *_dst, int16_t *_block, int stride){
70    int i;
71    pixel *dst = (pixel*)_dst;
72    dctcoef *block = (dctcoef*)_block;
73    stride >>= sizeof(pixel)-1;
74
75    block[0] += 32;
76
77    for( i = 0; i < 8; i++ )
78    {
79        const int a0 =  block[i+0*8] + block[i+4*8];
80        const int a2 =  block[i+0*8] - block[i+4*8];
81        const int a4 = (block[i+2*8]>>1) - block[i+6*8];
82        const int a6 = (block[i+6*8]>>1) + block[i+2*8];
83
84        const int b0 = a0 + a6;
85        const int b2 = a2 + a4;
86        const int b4 = a2 - a4;
87        const int b6 = a0 - a6;
88
89        const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
90        const int a3 =  block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
91        const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
92        const int a7 =  block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
93
94        const int b1 = (a7>>2) + a1;
95        const int b3 =  a3 + (a5>>2);
96        const int b5 = (a3>>2) - a5;
97        const int b7 =  a7 - (a1>>2);
98
99        block[i+0*8] = b0 + b7;
100        block[i+7*8] = b0 - b7;
101        block[i+1*8] = b2 + b5;
102        block[i+6*8] = b2 - b5;
103        block[i+2*8] = b4 + b3;
104        block[i+5*8] = b4 - b3;
105        block[i+3*8] = b6 + b1;
106        block[i+4*8] = b6 - b1;
107    }
108    for( i = 0; i < 8; i++ )
109    {
110        const int a0 =  block[0+i*8] + block[4+i*8];
111        const int a2 =  block[0+i*8] - block[4+i*8];
112        const int a4 = (block[2+i*8]>>1) - block[6+i*8];
113        const int a6 = (block[6+i*8]>>1) + block[2+i*8];
114
115        const int b0 = a0 + a6;
116        const int b2 = a2 + a4;
117        const int b4 = a2 - a4;
118        const int b6 = a0 - a6;
119
120        const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
121        const int a3 =  block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
122        const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
123        const int a7 =  block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
124
125        const int b1 = (a7>>2) + a1;
126        const int b3 =  a3 + (a5>>2);
127        const int b5 = (a3>>2) - a5;
128        const int b7 =  a7 - (a1>>2);
129
130        dst[i + 0*stride] = av_clip_pixel( dst[i + 0*stride] + ((b0 + b7) >> 6) );
131        dst[i + 1*stride] = av_clip_pixel( dst[i + 1*stride] + ((b2 + b5) >> 6) );
132        dst[i + 2*stride] = av_clip_pixel( dst[i + 2*stride] + ((b4 + b3) >> 6) );
133        dst[i + 3*stride] = av_clip_pixel( dst[i + 3*stride] + ((b6 + b1) >> 6) );
134        dst[i + 4*stride] = av_clip_pixel( dst[i + 4*stride] + ((b6 - b1) >> 6) );
135        dst[i + 5*stride] = av_clip_pixel( dst[i + 5*stride] + ((b4 - b3) >> 6) );
136        dst[i + 6*stride] = av_clip_pixel( dst[i + 6*stride] + ((b2 - b5) >> 6) );
137        dst[i + 7*stride] = av_clip_pixel( dst[i + 7*stride] + ((b0 - b7) >> 6) );
138    }
139
140    memset(block, 0, 64 * sizeof(dctcoef));
141}
142
143// assumes all AC coefs are 0
144void FUNCC(ff_h264_idct_dc_add)(uint8_t *_dst, int16_t *_block, int stride){
145    int i, j;
146    pixel *dst = (pixel*)_dst;
147    dctcoef *block = (dctcoef*)_block;
148    int dc = (block[0] + 32) >> 6;
149    stride /= sizeof(pixel);
150    block[0] = 0;
151    for( j = 0; j < 4; j++ )
152    {
153        for( i = 0; i < 4; i++ )
154            dst[i] = av_clip_pixel( dst[i] + dc );
155        dst += stride;
156    }
157}
158
159void FUNCC(ff_h264_idct8_dc_add)(uint8_t *_dst, int16_t *_block, int stride){
160    int i, j;
161    pixel *dst = (pixel*)_dst;
162    dctcoef *block = (dctcoef*)_block;
163    int dc = (block[0] + 32) >> 6;
164    block[0] = 0;
165    stride /= sizeof(pixel);
166    for( j = 0; j < 8; j++ )
167    {
168        for( i = 0; i < 8; i++ )
169            dst[i] = av_clip_pixel( dst[i] + dc );
170        dst += stride;
171    }
172}
173
174void FUNCC(ff_h264_idct_add16)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
175    int i;
176    for(i=0; i<16; i++){
177        int nnz = nnzc[ scan8[i] ];
178        if(nnz){
179            if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
180            else                                  FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
181        }
182    }
183}
184
185void FUNCC(ff_h264_idct_add16intra)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
186    int i;
187    for(i=0; i<16; i++){
188        if(nnzc[ scan8[i] ])             FUNCC(ff_h264_idct_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
189        else if(((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
190    }
191}
192
193void FUNCC(ff_h264_idct8_add4)(uint8_t *dst, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
194    int i;
195    for(i=0; i<16; i+=4){
196        int nnz = nnzc[ scan8[i] ];
197        if(nnz){
198            if(nnz==1 && ((dctcoef*)block)[i*16]) FUNCC(ff_h264_idct8_dc_add)(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
199            else                                  FUNCC(ff_h264_idct8_add   )(dst + block_offset[i], block + i*16*sizeof(pixel), stride);
200        }
201    }
202}
203
204void FUNCC(ff_h264_idct_add8)(uint8_t **dest, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
205    int i, j;
206    for(j=1; j<3; j++){
207        for(i=j*16; i<j*16+4; i++){
208            if(nnzc[ scan8[i] ])
209                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
210            else if(((dctcoef*)block)[i*16])
211                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
212        }
213    }
214}
215
216void FUNCC(ff_h264_idct_add8_422)(uint8_t **dest, const int *block_offset, int16_t *block, int stride, const uint8_t nnzc[15*8]){
217    int i, j;
218
219    for(j=1; j<3; j++){
220        for(i=j*16; i<j*16+4; i++){
221            if(nnzc[ scan8[i] ])
222                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
223            else if(((dctcoef*)block)[i*16])
224                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i], block + i*16*sizeof(pixel), stride);
225        }
226    }
227
228    for(j=1; j<3; j++){
229        for(i=j*16+4; i<j*16+8; i++){
230            if(nnzc[ scan8[i+4] ])
231                FUNCC(ff_h264_idct_add   )(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
232            else if(((dctcoef*)block)[i*16])
233                FUNCC(ff_h264_idct_dc_add)(dest[j-1] + block_offset[i+4], block + i*16*sizeof(pixel), stride);
234        }
235    }
236}
237
238/**
239 * IDCT transforms the 16 dc values and dequantizes them.
240 * @param qmul quantization parameter
241 */
242void FUNCC(ff_h264_luma_dc_dequant_idct)(int16_t *_output, int16_t *_input, int qmul){
243#define stride 16
244    int i;
245    int temp[16];
246    static const uint8_t x_offset[4]={0, 2*stride, 8*stride, 10*stride};
247    dctcoef *input = (dctcoef*)_input;
248    dctcoef *output = (dctcoef*)_output;
249
250    for(i=0; i<4; i++){
251        const int z0= input[4*i+0] + input[4*i+1];
252        const int z1= input[4*i+0] - input[4*i+1];
253        const int z2= input[4*i+2] - input[4*i+3];
254        const int z3= input[4*i+2] + input[4*i+3];
255
256        temp[4*i+0]= z0+z3;
257        temp[4*i+1]= z0-z3;
258        temp[4*i+2]= z1-z2;
259        temp[4*i+3]= z1+z2;
260    }
261
262    for(i=0; i<4; i++){
263        const int offset= x_offset[i];
264        const int z0= temp[4*0+i] + temp[4*2+i];
265        const int z1= temp[4*0+i] - temp[4*2+i];
266        const int z2= temp[4*1+i] - temp[4*3+i];
267        const int z3= temp[4*1+i] + temp[4*3+i];
268
269        output[stride* 0+offset]= ((((z0 + z3)*qmul + 128 ) >> 8));
270        output[stride* 1+offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
271        output[stride* 4+offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
272        output[stride* 5+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
273    }
274#undef stride
275}
276
277void FUNCC(ff_h264_chroma422_dc_dequant_idct)(int16_t *_block, int qmul){
278    const int stride= 16*2;
279    const int xStride= 16;
280    int i;
281    int temp[8];
282    static const uint8_t x_offset[2]={0, 16};
283    dctcoef *block = (dctcoef*)_block;
284
285    for(i=0; i<4; i++){
286        temp[2*i+0] = block[stride*i + xStride*0] + block[stride*i + xStride*1];
287        temp[2*i+1] = block[stride*i + xStride*0] - block[stride*i + xStride*1];
288    }
289
290    for(i=0; i<2; i++){
291        const int offset= x_offset[i];
292        const int z0= temp[2*0+i] + temp[2*2+i];
293        const int z1= temp[2*0+i] - temp[2*2+i];
294        const int z2= temp[2*1+i] - temp[2*3+i];
295        const int z3= temp[2*1+i] + temp[2*3+i];
296
297        block[stride*0+offset]= ((z0 + z3)*qmul + 128) >> 8;
298        block[stride*1+offset]= ((z1 + z2)*qmul + 128) >> 8;
299        block[stride*2+offset]= ((z1 - z2)*qmul + 128) >> 8;
300        block[stride*3+offset]= ((z0 - z3)*qmul + 128) >> 8;
301    }
302}
303
304void FUNCC(ff_h264_chroma_dc_dequant_idct)(int16_t *_block, int qmul){
305    const int stride= 16*2;
306    const int xStride= 16;
307    int a,b,c,d,e;
308    dctcoef *block = (dctcoef*)_block;
309
310    a= block[stride*0 + xStride*0];
311    b= block[stride*0 + xStride*1];
312    c= block[stride*1 + xStride*0];
313    d= block[stride*1 + xStride*1];
314
315    e= a-b;
316    a= a+b;
317    b= c-d;
318    c= c+d;
319
320    block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
321    block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
322    block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
323    block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
324}
325