Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > VSs > VSs__H264__App

comparison libavcodec/h264_idct.c @ 9:ea1ba68cf0ed

find changesets by author, revision, files, or words in the commit message
update to match api changes + add sscc produced source
author Nina Engelhardt <nengel@mailbox.tu-berlin.de>
date Wed, 05 Jun 2013 14:43:26 +0200
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:26ba90c7d140
1 /*
2 * H.264 IDCT
3 * Copyright (c) 2004 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 "dsputil.h"
29 #include "h264_data.h"
30
31 static av_always_inline void idct_internal(uint8_t *dst, DCTELEM *block, int stride, int block_stride, int shift, int add){
32 int i;
33 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
34
35 block[0] += 1<<(shift-1);
36
37 for(i=0; i<4; i++){
38 const int z0= block[0 + block_stride*i] + block[2 + block_stride*i];
39 const int z1= block[0 + block_stride*i] - block[2 + block_stride*i];
40 const int z2= (block[1 + block_stride*i]>>1) - block[3 + block_stride*i];
41 const int z3= block[1 + block_stride*i] + (block[3 + block_stride*i]>>1);
42
43 block[0 + block_stride*i]= z0 + z3;
44 block[1 + block_stride*i]= z1 + z2;
45 block[2 + block_stride*i]= z1 - z2;
46 block[3 + block_stride*i]= z0 - z3;
47 }
48
49 for(i=0; i<4; i++){
50 const int z0= block[i + block_stride*0] + block[i + block_stride*2];
51 const int z1= block[i + block_stride*0] - block[i + block_stride*2];
52 const int z2= (block[i + block_stride*1]>>1) - block[i + block_stride*3];
53 const int z3= block[i + block_stride*1] + (block[i + block_stride*3]>>1);
54
55 dst[i + 0*stride]= cm[ add*dst[i + 0*stride] + ((z0 + z3) >> shift) ];
56 dst[i + 1*stride]= cm[ add*dst[i + 1*stride] + ((z1 + z2) >> shift) ];
57 dst[i + 2*stride]= cm[ add*dst[i + 2*stride] + ((z1 - z2) >> shift) ];
58 dst[i + 3*stride]= cm[ add*dst[i + 3*stride] + ((z0 - z3) >> shift) ];
59 }
60 }
61
62 void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride){
63 idct_internal(dst, block, stride, 4, 6, 1);
64 }
65
66 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block){
67 idct_internal(dst, block, stride, 8, 3, 1);
68 }
69
70 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block){
71 idct_internal(dst, block, stride, 8, 3, 0);
72 }
73
74 void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride){
75 int i;
76 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
77
78 block[0] += 32;
79
80 for( i = 0; i < 8; i++ )
81 {
82 const int a0 = block[0+i*8] + block[4+i*8];
83 const int a2 = block[0+i*8] - block[4+i*8];
84 const int a4 = (block[2+i*8]>>1) - block[6+i*8];
85 const int a6 = (block[6+i*8]>>1) + block[2+i*8];
86
87 const int b0 = a0 + a6;
88 const int b2 = a2 + a4;
89 const int b4 = a2 - a4;
90 const int b6 = a0 - a6;
91
92 const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
93 const int a3 = block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
94 const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
95 const int a7 = block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
96
97 const int b1 = (a7>>2) + a1;
98 const int b3 = a3 + (a5>>2);
99 const int b5 = (a3>>2) - a5;
100 const int b7 = a7 - (a1>>2);
101
102 block[0+i*8] = b0 + b7;
103 block[7+i*8] = b0 - b7;
104 block[1+i*8] = b2 + b5;
105 block[6+i*8] = b2 - b5;
106 block[2+i*8] = b4 + b3;
107 block[5+i*8] = b4 - b3;
108 block[3+i*8] = b6 + b1;
109 block[4+i*8] = b6 - b1;
110 }
111 for( i = 0; i < 8; i++ )
112 {
113 const int a0 = block[i+0*8] + block[i+4*8];
114 const int a2 = block[i+0*8] - block[i+4*8];
115 const int a4 = (block[i+2*8]>>1) - block[i+6*8];
116 const int a6 = (block[i+6*8]>>1) + block[i+2*8];
117
118 const int b0 = a0 + a6;
119 const int b2 = a2 + a4;
120 const int b4 = a2 - a4;
121 const int b6 = a0 - a6;
122
123 const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
124 const int a3 = block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
125 const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
126 const int a7 = block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
127
128 const int b1 = (a7>>2) + a1;
129 const int b3 = a3 + (a5>>2);
130 const int b5 = (a3>>2) - a5;
131 const int b7 = a7 - (a1>>2);
132
133 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b7) >> 6) ];
134 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b2 + b5) >> 6) ];
135 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b4 + b3) >> 6) ];
136 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b6 + b1) >> 6) ];
137 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b6 - b1) >> 6) ];
138 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b4 - b3) >> 6) ];
139 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b2 - b5) >> 6) ];
140 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b7) >> 6) ];
141 }
142 }
143
144 // assumes all AC coefs are 0
145 void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
146 int i, j;
147 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
148 int dc = (block[0] + 32) >> 6;
149 for( j = 0; j < 4; j++ )
150 {
151 for( i = 0; i < 4; i++ )
152 dst[i] = cm[ dst[i] + dc ];
153 dst += stride;
154 }
155 }
156
157 void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
158 int i, j;
159 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
160 int dc = (block[0] + 32) >> 6;
161 for( j = 0; j < 8; j++ )
162 {
163 for( i = 0; i < 8; i++ )
164 dst[i] = cm[ dst[i] + dc ];
165 dst += stride;
166 }
167 }
168
169 void ff_h264_idct_add16_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
170 int i;
171 for(i=0; i<16; i++){
172 int nnz = nnzc[ scan8[i] ];
173 if(nnz){
174 if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
175 else idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
176 }
177 }
178 }
179
180 void ff_h264_idct_add16intra_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
181 int i;
182 for(i=0; i<16; i++){
183 if(nnzc[ scan8[i] ]) idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
184 else if(block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
185 }
186 }
187
188 void ff_h264_idct8_add4_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
189 int i;
190 for(i=0; i<16; i+=4){
191 int nnz = nnzc[ scan8[i] ];
192 if(nnz){
193 if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_c(dst + block_offset[i], block + i*16, stride);
194 else ff_h264_idct8_add_c (dst + block_offset[i], block + i*16, stride);
195 }
196 }
197 }
198
199 void ff_h264_idct_add8_c(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
200 int i;
201 for(i=16; i<16+8; i++){
202 if(nnzc[ scan8[i] ])
203 ff_h264_idct_add_c (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
204 else if(block[i*16])
205 ff_h264_idct_dc_add_c(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
206 }
207 }
208
209 /**
210 * IDCT transforms the 16 dc values and dequantizes them.
211 * @param qp quantization parameter
212 */
213 void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qmul){
214 #define stride 16
215 int i;
216 int temp[16]; //FIXME check if this is a good idea
217 static const int x_offset[4]={0, 1*stride, 4* stride, 5*stride};
218 static const int y_offset[4]={0, 2*stride, 8* stride, 10*stride};
219
220 //return;
221 for(i=0; i<4; i++){
222 const int offset= y_offset[i];
223 const int z0= block[offset+stride*0] + block[offset+stride*4];
224 const int z1= block[offset+stride*0] - block[offset+stride*4];
225 const int z2= block[offset+stride*1] - block[offset+stride*5];
226 const int z3= block[offset+stride*1] + block[offset+stride*5];
227
228 temp[4*i+0]= z0+z3;
229 temp[4*i+1]= z1+z2;
230 temp[4*i+2]= z1-z2;
231 temp[4*i+3]= z0-z3;
232 }
233
234 for(i=0; i<4; i++){
235 const int offset= x_offset[i];
236 const int z0= temp[4*0+i] + temp[4*2+i];
237 const int z1= temp[4*0+i] - temp[4*2+i];
238 const int z2= temp[4*1+i] - temp[4*3+i];
239 const int z3= temp[4*1+i] + temp[4*3+i];
240
241 block[stride*0 +offset]= ((((z0 + z3)*qmul + 128 ) >> 8)); //FIXME think about merging this into decode_residual
242 block[stride*2 +offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
243 block[stride*8 +offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
244 block[stride*10+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
245 }
246 }
247
248 #undef xStride
249 #undef stride
250
251 void chroma_dc_dequant_idct_c(DCTELEM *block, int qmul){
252 const int stride= 16*2;
253 const int xStride= 16;
254 int a,b,c,d,e;
255
256 a= block[stride*0 + xStride*0];
257 b= block[stride*0 + xStride*1];
258 c= block[stride*1 + xStride*0];
259 d= block[stride*1 + xStride*1];
260
261 e= a-b;
262 a= a+b;
263 b= c-d;
264 c= c+d;
265
266 block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
267 block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
268 block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
269 block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
270 }