Mercurial > libavcodec.hg

comparison vp3dsp.c @ 2693:02925a3903b6 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
porting vp3 idct over to lavc idct api
author michael
date Tue, 17 May 2005 11:12:04 +0000
parents fb8f26fc57eb
children fd5d7c732c6b
comparison
equal deleted inserted replaced
2692:fb8f26fc57eb 2693:02925a3903b6
34 #define xC4S4 46341 34 #define xC4S4 46341
35 #define xC5S3 36410 35 #define xC5S3 36410
36 #define xC6S2 25080 36 #define xC6S2 25080
37 #define xC7S1 12785 37 #define xC7S1 12785
38 38
39 void vp3_dsp_init_c(void) 39 static always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
40 { 40 {
41 /* nop */ 41 int16_t *ip = input;
42 } 42 uint8_t *cm = cropTbl + MAX_NEG_CROP;
43 43
44 void vp3_idct_c(int16_t *input_data, int16_t *dequant_matrix, 44 int A_, B_, C_, D_, _Ad, _Bd, _Cd, _Dd, E_, F_, G_, H_;
45 int coeff_count, int16_t *output_data) 45 int _Ed, _Gd, _Add, _Bdd, _Fd, _Hd;
46 { 46 int t1, t2;
47 int32_t dequantized_data[64];
48 int32_t *ip = dequantized_data;
49 int16_t *op = output_data;
50
51 int32_t A_, B_, C_, D_, _Ad, _Bd, _Cd, _Dd, E_, F_, G_, H_;
52 int32_t _Ed, _Gd, _Add, _Bdd, _Fd, _Hd;
53 int32_t t1, t2;
54 47
55 int i, j; 48 int i, j;
56 49
57 /* de-zigzag and dequantize */
58 for (i = 0; i < coeff_count; i++) {
59 dequantized_data[i] = dequant_matrix[i] * input_data[i];
60 }
61
62 /* Inverse DCT on the rows now */ 50 /* Inverse DCT on the rows now */
63 for (i = 0; i < 8; i++) { 51 for (i = 0; i < 8; i++) {
64 /* Check for non-zero values */ 52 /* Check for non-zero values */
65 if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) { 53 if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
66 t1 = (int32_t)(xC1S7 * ip[1]); 54 t1 = (int32_t)(xC1S7 * ip[1]);
129 117
130 _Fd = F_ - _Ad; 118 _Fd = F_ - _Ad;
131 _Hd = _Bd + H_; 119 _Hd = _Bd + H_;
132 120
133 /* Final sequence of operations over-write original inputs. */ 121 /* Final sequence of operations over-write original inputs. */
134 ip[0] = (int16_t)((_Gd + _Cd ) >> 0); 122 ip[0] = _Gd + _Cd ;
135 ip[7] = (int16_t)((_Gd - _Cd ) >> 0); 123 ip[7] = _Gd - _Cd ;
136 124
137 ip[1] = (int16_t)((_Add + _Hd ) >> 0); 125 ip[1] = _Add + _Hd;
138 ip[2] = (int16_t)((_Add - _Hd ) >> 0); 126 ip[2] = _Add - _Hd;
139 127
140 ip[3] = (int16_t)((_Ed + _Dd ) >> 0); 128 ip[3] = _Ed + _Dd ;
141 ip[4] = (int16_t)((_Ed - _Dd ) >> 0); 129 ip[4] = _Ed - _Dd ;
142 130
143 ip[5] = (int16_t)((_Fd + _Bdd ) >> 0); 131 ip[5] = _Fd + _Bdd;
144 ip[6] = (int16_t)((_Fd - _Bdd ) >> 0); 132 ip[6] = _Fd - _Bdd;
145 133
146 } 134 }
147 135
148 ip += 8; /* next row */ 136 ip += 8; /* next row */
149 } 137 }
150 138
151 ip = dequantized_data; 139 ip = input;
152 140
153 for ( i = 0; i < 8; i++) { 141 for ( i = 0; i < 8; i++) {
154 /* Check for non-zero values (bitwise or faster than ||) */ 142 /* Check for non-zero values (bitwise or faster than ||) */
155 if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] | 143 if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
156 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) { 144 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
157 145
158 t1 = (int32_t)(xC1S7 * ip[1*8]); 146 t1 = (int32_t)(xC1S7 * ip[1*8]);
159 t2 = (int32_t)(xC7S1 * ip[7*8]); 147 t2 = (int32_t)(xC7S1 * ip[7*8]);
160 t1 >>= 16; 148 t1 >>= 16;
220 _Bdd = _Bd - H_; 208 _Bdd = _Bd - H_;
221 209
222 _Fd = F_ - _Ad; 210 _Fd = F_ - _Ad;
223 _Hd = _Bd + H_; 211 _Hd = _Bd + H_;
224 212
213 if(type==1){ //HACK
214 _Gd += 16*128;
215 _Add+= 16*128;
216 _Ed += 16*128;
217 _Fd += 16*128;
218 }
225 _Gd += IdctAdjustBeforeShift; 219 _Gd += IdctAdjustBeforeShift;
226 _Add += IdctAdjustBeforeShift; 220 _Add += IdctAdjustBeforeShift;
227 _Ed += IdctAdjustBeforeShift; 221 _Ed += IdctAdjustBeforeShift;
228 _Fd += IdctAdjustBeforeShift; 222 _Fd += IdctAdjustBeforeShift;
229 223
230 /* Final sequence of operations over-write original inputs. */ 224 /* Final sequence of operations over-write original inputs. */
231 op[0*8] = (int16_t)((_Gd + _Cd ) >> 4); 225 if(type==0){
232 op[7*8] = (int16_t)((_Gd - _Cd ) >> 4); 226 ip[0*8] = (_Gd + _Cd ) >> 4;
233 227 ip[7*8] = (_Gd - _Cd ) >> 4;
234 op[1*8] = (int16_t)((_Add + _Hd ) >> 4); 228
235 op[2*8] = (int16_t)((_Add - _Hd ) >> 4); 229 ip[1*8] = (_Add + _Hd ) >> 4;
236 230 ip[2*8] = (_Add - _Hd ) >> 4;
237 op[3*8] = (int16_t)((_Ed + _Dd ) >> 4); 231
238 op[4*8] = (int16_t)((_Ed - _Dd ) >> 4); 232 ip[3*8] = (_Ed + _Dd ) >> 4;
239 233 ip[4*8] = (_Ed - _Dd ) >> 4;
240 op[5*8] = (int16_t)((_Fd + _Bdd ) >> 4); 234
241 op[6*8] = (int16_t)((_Fd - _Bdd ) >> 4); 235 ip[5*8] = (_Fd + _Bdd ) >> 4;
236 ip[6*8] = (_Fd - _Bdd ) >> 4;
237 }else if(type==1){
238 dst[0*stride] = cm[(_Gd + _Cd ) >> 4];
239 dst[7*stride] = cm[(_Gd - _Cd ) >> 4];
240
241 dst[1*stride] = cm[(_Add + _Hd ) >> 4];
242 dst[2*stride] = cm[(_Add - _Hd ) >> 4];
243
244 dst[3*stride] = cm[(_Ed + _Dd ) >> 4];
245 dst[4*stride] = cm[(_Ed - _Dd ) >> 4];
246
247 dst[5*stride] = cm[(_Fd + _Bdd ) >> 4];
248 dst[6*stride] = cm[(_Fd - _Bdd ) >> 4];
249 }else{
250 dst[0*stride] = cm[dst[0*stride] + ((_Gd + _Cd ) >> 4)];
251 dst[7*stride] = cm[dst[7*stride] + ((_Gd - _Cd ) >> 4)];
252
253 dst[1*stride] = cm[dst[1*stride] + ((_Add + _Hd ) >> 4)];
254 dst[2*stride] = cm[dst[2*stride] + ((_Add - _Hd ) >> 4)];
255
256 dst[3*stride] = cm[dst[3*stride] + ((_Ed + _Dd ) >> 4)];
257 dst[4*stride] = cm[dst[4*stride] + ((_Ed - _Dd ) >> 4)];
258
259 dst[5*stride] = cm[dst[5*stride] + ((_Fd + _Bdd ) >> 4)];
260 dst[6*stride] = cm[dst[6*stride] + ((_Fd - _Bdd ) >> 4)];
261 }
242 262
243 } else { 263 } else {
244 264 if(type==0){
245 op[0*8] = 0; 265 ip[0*8] =
246 op[7*8] = 0; 266 ip[1*8] =
247 op[1*8] = 0; 267 ip[2*8] =
248 op[2*8] = 0; 268 ip[3*8] =
249 op[3*8] = 0; 269 ip[4*8] =
250 op[4*8] = 0; 270 ip[5*8] =
251 op[5*8] = 0; 271 ip[6*8] =
252 op[6*8] = 0; 272 ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
273 }else if(type==1){
274 dst[0*stride]=
275 dst[1*stride]=
276 dst[2*stride]=
277 dst[3*stride]=
278 dst[4*stride]=
279 dst[5*stride]=
280 dst[6*stride]=
281 dst[7*stride]= 128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
282 }else{
283 if(ip[0*8]){
284 int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
285 dst[0*stride] = cm[dst[0*stride] + v];
286 dst[1*stride] = cm[dst[1*stride] + v];
287 dst[2*stride] = cm[dst[2*stride] + v];
288 dst[3*stride] = cm[dst[3*stride] + v];
289 dst[4*stride] = cm[dst[4*stride] + v];
290 dst[5*stride] = cm[dst[5*stride] + v];
291 dst[6*stride] = cm[dst[6*stride] + v];
292 dst[7*stride] = cm[dst[7*stride] + v];
293 }
294 }
253 } 295 }
254 296
255 ip++; /* next column */ 297 ip++; /* next column */
256 op++; 298 dst++;
257 } 299 }
258 } 300 }
301
302 void ff_vp3_idct_c(DCTELEM *block/* align 16*/){
303 idct(NULL, 0, block, 0);
304 }
305
306 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
307 idct(dest, line_size, block, 1);
308 }
309
310 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
311 idct(dest, line_size, block, 2);
312 }