flatten an array, since gcc fails at optimizing multidimensional arrays
[libav.git] / libavcodec / h264idct.c
CommitLineData
0fa8158d
MN
1/*
2 * H.264 IDCT
3 * Copyright (c) 2004 Michael Niedermayer <michaelni@gmx.at>
4 *
b78e7197
DB
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
0fa8158d
MN
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
b78e7197 10 * version 2.1 of the License, or (at your option) any later version.
0fa8158d 11 *
b78e7197 12 * FFmpeg is distributed in the hope that it will be useful,
0fa8158d
MN
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
b78e7197 18 * License along with FFmpeg; if not, write to the Free Software
5509bffa 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
0fa8158d 20 */
115329f1 21
0fa8158d
MN
22/**
23 * @file h264-idct.c
24 * H.264 IDCT.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
115329f1 27
0fa8158d
MN
28#include "dsputil.h"
29
849f1035 30static av_always_inline void idct_internal(uint8_t *dst, DCTELEM *block, int stride, int block_stride, int shift, int add){
0fa8158d 31 int i;
55fde95e 32 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
0fa8158d
MN
33
34 block[0] += 1<<(shift-1);
35
36 for(i=0; i<4; i++){
37 const int z0= block[0 + block_stride*i] + block[2 + block_stride*i];
38 const int z1= block[0 + block_stride*i] - block[2 + block_stride*i];
39 const int z2= (block[1 + block_stride*i]>>1) - block[3 + block_stride*i];
40 const int z3= block[1 + block_stride*i] + (block[3 + block_stride*i]>>1);
41
42 block[0 + block_stride*i]= z0 + z3;
43 block[1 + block_stride*i]= z1 + z2;
44 block[2 + block_stride*i]= z1 - z2;
45 block[3 + block_stride*i]= z0 - z3;
46 }
47
48 for(i=0; i<4; i++){
49 const int z0= block[i + block_stride*0] + block[i + block_stride*2];
50 const int z1= block[i + block_stride*0] - block[i + block_stride*2];
51 const int z2= (block[i + block_stride*1]>>1) - block[i + block_stride*3];
52 const int z3= block[i + block_stride*1] + (block[i + block_stride*3]>>1);
53
54 dst[i + 0*stride]= cm[ add*dst[i + 0*stride] + ((z0 + z3) >> shift) ];
55 dst[i + 1*stride]= cm[ add*dst[i + 1*stride] + ((z1 + z2) >> shift) ];
56 dst[i + 2*stride]= cm[ add*dst[i + 2*stride] + ((z1 - z2) >> shift) ];
57 dst[i + 3*stride]= cm[ add*dst[i + 3*stride] + ((z0 - z3) >> shift) ];
58 }
59}
60
61void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride){
62 idct_internal(dst, block, stride, 4, 6, 1);
63}
64
65void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block){
66 idct_internal(dst, block, stride, 8, 3, 1);
67}
68
69void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block){
70 idct_internal(dst, block, stride, 8, 3, 0);
71}
43efd19a
LM
72
73void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride){
74 int i;
55fde95e 75 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
43efd19a
LM
76
77 block[0] += 32;
78
79 for( i = 0; i < 8; i++ )
80 {
1cca8d24
LM
81 const int a0 = block[0+i*8] + block[4+i*8];
82 const int a2 = block[0+i*8] - block[4+i*8];
83 const int a4 = (block[2+i*8]>>1) - block[6+i*8];
84 const int a6 = (block[6+i*8]>>1) + block[2+i*8];
43efd19a
LM
85
86 const int b0 = a0 + a6;
87 const int b2 = a2 + a4;
88 const int b4 = a2 - a4;
89 const int b6 = a0 - a6;
90
1cca8d24
LM
91 const int a1 = -block[3+i*8] + block[5+i*8] - block[7+i*8] - (block[7+i*8]>>1);
92 const int a3 = block[1+i*8] + block[7+i*8] - block[3+i*8] - (block[3+i*8]>>1);
93 const int a5 = -block[1+i*8] + block[7+i*8] + block[5+i*8] + (block[5+i*8]>>1);
94 const int a7 = block[3+i*8] + block[5+i*8] + block[1+i*8] + (block[1+i*8]>>1);
43efd19a
LM
95
96 const int b1 = (a7>>2) + a1;
97 const int b3 = a3 + (a5>>2);
98 const int b5 = (a3>>2) - a5;
99 const int b7 = a7 - (a1>>2);
100
1cca8d24
LM
101 block[0+i*8] = b0 + b7;
102 block[7+i*8] = b0 - b7;
103 block[1+i*8] = b2 + b5;
104 block[6+i*8] = b2 - b5;
105 block[2+i*8] = b4 + b3;
106 block[5+i*8] = b4 - b3;
107 block[3+i*8] = b6 + b1;
108 block[4+i*8] = b6 - b1;
43efd19a
LM
109 }
110 for( i = 0; i < 8; i++ )
111 {
1cca8d24
LM
112 const int a0 = block[i+0*8] + block[i+4*8];
113 const int a2 = block[i+0*8] - block[i+4*8];
114 const int a4 = (block[i+2*8]>>1) - block[i+6*8];
115 const int a6 = (block[i+6*8]>>1) + block[i+2*8];
43efd19a
LM
116
117 const int b0 = a0 + a6;
118 const int b2 = a2 + a4;
119 const int b4 = a2 - a4;
120 const int b6 = a0 - a6;
121
1cca8d24
LM
122 const int a1 = -block[i+3*8] + block[i+5*8] - block[i+7*8] - (block[i+7*8]>>1);
123 const int a3 = block[i+1*8] + block[i+7*8] - block[i+3*8] - (block[i+3*8]>>1);
124 const int a5 = -block[i+1*8] + block[i+7*8] + block[i+5*8] + (block[i+5*8]>>1);
125 const int a7 = block[i+3*8] + block[i+5*8] + block[i+1*8] + (block[i+1*8]>>1);
43efd19a
LM
126
127 const int b1 = (a7>>2) + a1;
128 const int b3 = a3 + (a5>>2);
129 const int b5 = (a3>>2) - a5;
130 const int b7 = a7 - (a1>>2);
131
132 dst[i + 0*stride] = cm[ dst[i + 0*stride] + ((b0 + b7) >> 6) ];
133 dst[i + 1*stride] = cm[ dst[i + 1*stride] + ((b2 + b5) >> 6) ];
134 dst[i + 2*stride] = cm[ dst[i + 2*stride] + ((b4 + b3) >> 6) ];
135 dst[i + 3*stride] = cm[ dst[i + 3*stride] + ((b6 + b1) >> 6) ];
136 dst[i + 4*stride] = cm[ dst[i + 4*stride] + ((b6 - b1) >> 6) ];
137 dst[i + 5*stride] = cm[ dst[i + 5*stride] + ((b4 - b3) >> 6) ];
138 dst[i + 6*stride] = cm[ dst[i + 6*stride] + ((b2 - b5) >> 6) ];
139 dst[i + 7*stride] = cm[ dst[i + 7*stride] + ((b0 - b7) >> 6) ];
140 }
141}
ef9d1d15
LM
142
143// assumes all AC coefs are 0
144void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
145 int i, j;
55fde95e 146 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
ef9d1d15
LM
147 int dc = (block[0] + 32) >> 6;
148 for( j = 0; j < 4; j++ )
149 {
150 for( i = 0; i < 4; i++ )
151 dst[i] = cm[ dst[i] + dc ];
152 dst += stride;
153 }
154}
155
156void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride){
157 int i, j;
55fde95e 158 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
ef9d1d15
LM
159 int dc = (block[0] + 32) >> 6;
160 for( j = 0; j < 8; j++ )
161 {
162 for( i = 0; i < 8; i++ )
163 dst[i] = cm[ dst[i] + dc ];
164 dst += stride;
165 }
166}
ac223859
MN
167
168//FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
169static const uint8_t scan8[16 + 2*4]={
170 4+1*8, 5+1*8, 4+2*8, 5+2*8,
171 6+1*8, 7+1*8, 6+2*8, 7+2*8,
172 4+3*8, 5+3*8, 4+4*8, 5+4*8,
173 6+3*8, 7+3*8, 6+4*8, 7+4*8,
174 1+1*8, 2+1*8,
175 1+2*8, 2+2*8,
176 1+4*8, 2+4*8,
177 1+5*8, 2+5*8,
178};
179
180void ff_h264_idct_add16_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 int nnz = nnzc[ scan8[i] ];
184 if(nnz){
185 if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
186 else idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
187 }
188 }
189}
190
191void ff_h264_idct_add16intra_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
192 int i;
193 for(i=0; i<16; i++){
194 if(nnzc[ scan8[i] ]) idct_internal (dst + block_offset[i], block + i*16, stride, 4, 6, 1);
195 else if(block[i*16]) ff_h264_idct_dc_add_c(dst + block_offset[i], block + i*16, stride);
196 }
197}
198
199void ff_h264_idct8_add4_c(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
200 int i;
201 for(i=0; i<16; i+=4){
202 int nnz = nnzc[ scan8[i] ];
203 if(nnz){
204 if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_c(dst + block_offset[i], block + i*16, stride);
205 else ff_h264_idct8_add_c (dst + block_offset[i], block + i*16, stride);
206 }
207 }
208}
209
210void ff_h264_idct_add8_c(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
211 int i;
212 for(i=16; i<16+8; i++){
213 if(nnzc[ scan8[i] ])
214 ff_h264_idct_add_c (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
215 else if(block[i*16])
216 ff_h264_idct_dc_add_c(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
217 }
218}