vp3: DC-only IDCT
[libav.git] / libavcodec / vp3dsp.c
1 /*
2 * Copyright (C) 2004 the ffmpeg project
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file libavcodec/vp3dsp.c
23 * Standard C DSP-oriented functions cribbed from the original VP3
24 * source code.
25 */
26
27 #include "avcodec.h"
28 #include "dsputil.h"
29
30 #define IdctAdjustBeforeShift 8
31 #define xC1S7 64277
32 #define xC2S6 60547
33 #define xC3S5 54491
34 #define xC4S4 46341
35 #define xC5S3 36410
36 #define xC6S2 25080
37 #define xC7S1 12785
38
39 #define M(a,b) (((a) * (b))>>16)
40
41 static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
42 {
43 int16_t *ip = input;
44 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
45
46 int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
47 int Ed, Gd, Add, Bdd, Fd, Hd;
48
49 int i;
50
51 /* Inverse DCT on the rows now */
52 for (i = 0; i < 8; i++) {
53 /* Check for non-zero values */
54 if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
55 A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
56 B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
57 C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
58 D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
59
60 Ad = M(xC4S4, (A - C));
61 Bd = M(xC4S4, (B - D));
62
63 Cd = A + C;
64 Dd = B + D;
65
66 E = M(xC4S4, (ip[0] + ip[4]));
67 F = M(xC4S4, (ip[0] - ip[4]));
68
69 G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
70 H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
71
72 Ed = E - G;
73 Gd = E + G;
74
75 Add = F + Ad;
76 Bdd = Bd - H;
77
78 Fd = F - Ad;
79 Hd = Bd + H;
80
81 /* Final sequence of operations over-write original inputs. */
82 ip[0] = Gd + Cd ;
83 ip[7] = Gd - Cd ;
84
85 ip[1] = Add + Hd;
86 ip[2] = Add - Hd;
87
88 ip[3] = Ed + Dd ;
89 ip[4] = Ed - Dd ;
90
91 ip[5] = Fd + Bdd;
92 ip[6] = Fd - Bdd;
93 }
94
95 ip += 8; /* next row */
96 }
97
98 ip = input;
99
100 for ( i = 0; i < 8; i++) {
101 /* Check for non-zero values (bitwise or faster than ||) */
102 if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
103 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
104
105 A = M(xC1S7, ip[1*8]) + M(xC7S1, ip[7*8]);
106 B = M(xC7S1, ip[1*8]) - M(xC1S7, ip[7*8]);
107 C = M(xC3S5, ip[3*8]) + M(xC5S3, ip[5*8]);
108 D = M(xC3S5, ip[5*8]) - M(xC5S3, ip[3*8]);
109
110 Ad = M(xC4S4, (A - C));
111 Bd = M(xC4S4, (B - D));
112
113 Cd = A + C;
114 Dd = B + D;
115
116 E = M(xC4S4, (ip[0*8] + ip[4*8])) + 8;
117 F = M(xC4S4, (ip[0*8] - ip[4*8])) + 8;
118
119 if(type==1){ //HACK
120 E += 16*128;
121 F += 16*128;
122 }
123
124 G = M(xC2S6, ip[2*8]) + M(xC6S2, ip[6*8]);
125 H = M(xC6S2, ip[2*8]) - M(xC2S6, ip[6*8]);
126
127 Ed = E - G;
128 Gd = E + G;
129
130 Add = F + Ad;
131 Bdd = Bd - H;
132
133 Fd = F - Ad;
134 Hd = Bd + H;
135
136 /* Final sequence of operations over-write original inputs. */
137 if(type==0){
138 ip[0*8] = (Gd + Cd ) >> 4;
139 ip[7*8] = (Gd - Cd ) >> 4;
140
141 ip[1*8] = (Add + Hd ) >> 4;
142 ip[2*8] = (Add - Hd ) >> 4;
143
144 ip[3*8] = (Ed + Dd ) >> 4;
145 ip[4*8] = (Ed - Dd ) >> 4;
146
147 ip[5*8] = (Fd + Bdd ) >> 4;
148 ip[6*8] = (Fd - Bdd ) >> 4;
149 }else if(type==1){
150 dst[0*stride] = cm[(Gd + Cd ) >> 4];
151 dst[7*stride] = cm[(Gd - Cd ) >> 4];
152
153 dst[1*stride] = cm[(Add + Hd ) >> 4];
154 dst[2*stride] = cm[(Add - Hd ) >> 4];
155
156 dst[3*stride] = cm[(Ed + Dd ) >> 4];
157 dst[4*stride] = cm[(Ed - Dd ) >> 4];
158
159 dst[5*stride] = cm[(Fd + Bdd ) >> 4];
160 dst[6*stride] = cm[(Fd - Bdd ) >> 4];
161 }else{
162 dst[0*stride] = cm[dst[0*stride] + ((Gd + Cd ) >> 4)];
163 dst[7*stride] = cm[dst[7*stride] + ((Gd - Cd ) >> 4)];
164
165 dst[1*stride] = cm[dst[1*stride] + ((Add + Hd ) >> 4)];
166 dst[2*stride] = cm[dst[2*stride] + ((Add - Hd ) >> 4)];
167
168 dst[3*stride] = cm[dst[3*stride] + ((Ed + Dd ) >> 4)];
169 dst[4*stride] = cm[dst[4*stride] + ((Ed - Dd ) >> 4)];
170
171 dst[5*stride] = cm[dst[5*stride] + ((Fd + Bdd ) >> 4)];
172 dst[6*stride] = cm[dst[6*stride] + ((Fd - Bdd ) >> 4)];
173 }
174
175 } else {
176 if(type==0){
177 ip[0*8] =
178 ip[1*8] =
179 ip[2*8] =
180 ip[3*8] =
181 ip[4*8] =
182 ip[5*8] =
183 ip[6*8] =
184 ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
185 }else if(type==1){
186 dst[0*stride]=
187 dst[1*stride]=
188 dst[2*stride]=
189 dst[3*stride]=
190 dst[4*stride]=
191 dst[5*stride]=
192 dst[6*stride]=
193 dst[7*stride]= cm[128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20)];
194 }else{
195 if(ip[0*8]){
196 int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
197 dst[0*stride] = cm[dst[0*stride] + v];
198 dst[1*stride] = cm[dst[1*stride] + v];
199 dst[2*stride] = cm[dst[2*stride] + v];
200 dst[3*stride] = cm[dst[3*stride] + v];
201 dst[4*stride] = cm[dst[4*stride] + v];
202 dst[5*stride] = cm[dst[5*stride] + v];
203 dst[6*stride] = cm[dst[6*stride] + v];
204 dst[7*stride] = cm[dst[7*stride] + v];
205 }
206 }
207 }
208
209 ip++; /* next column */
210 dst++;
211 }
212 }
213
214 void ff_vp3_idct_c(DCTELEM *block/* align 16*/){
215 idct(NULL, 0, block, 0);
216 }
217
218 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
219 idct(dest, line_size, block, 1);
220 }
221
222 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
223 idct(dest, line_size, block, 2);
224 }
225
226 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/){
227 const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
228 int i, dc = block[0];
229 dc = (46341*dc)>>16;
230 dc = (46341*dc + (8<<16))>>20;
231
232 for(i = 0; i < 8; i++){
233 dest[0] = cm[dest[0]+dc];
234 dest[1] = cm[dest[1]+dc];
235 dest[2] = cm[dest[2]+dc];
236 dest[3] = cm[dest[3]+dc];
237 dest[4] = cm[dest[4]+dc];
238 dest[5] = cm[dest[5]+dc];
239 dest[6] = cm[dest[6]+dc];
240 dest[7] = cm[dest[7]+dc];
241 dest += line_size;
242 }
243 }
244
245 void ff_vp3_v_loop_filter_c(uint8_t *first_pixel, int stride, int *bounding_values)
246 {
247 unsigned char *end;
248 int filter_value;
249 const int nstride= -stride;
250
251 for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
252 filter_value =
253 (first_pixel[2 * nstride] - first_pixel[ stride])
254 +3*(first_pixel[0 ] - first_pixel[nstride]);
255 filter_value = bounding_values[(filter_value + 4) >> 3];
256 first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
257 first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
258 }
259 }
260
261 void ff_vp3_h_loop_filter_c(uint8_t *first_pixel, int stride, int *bounding_values)
262 {
263 unsigned char *end;
264 int filter_value;
265
266 for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
267 filter_value =
268 (first_pixel[-2] - first_pixel[ 1])
269 +3*(first_pixel[ 0] - first_pixel[-1]);
270 filter_value = bounding_values[(filter_value + 4) >> 3];
271 first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
272 first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
273 }
274 }