Mercurial > libavcodec.hg

comparison simple_idct.c @ 476:ec13b0a726c3 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
removed unused code - began to merge alpha specific stuff - added mac macros for suitable CPUs
author bellard
date Wed, 05 Jun 2002 14:25:06 +0000
parents 9b73bce5071a
children 40ffce2cb6ef
comparison
equal deleted inserted replaced
475:6ccc48e4c41b 476:ec13b0a726c3
45 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 45 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
46 #define ROW_SHIFT 11 46 #define ROW_SHIFT 11
47 #define COL_SHIFT 20 // 6 47 #define COL_SHIFT 20 // 6
48 #endif 48 #endif
49 49
50 /* 8x8 Matrix used to do a trivial (slow) 8 point IDCT */ 50 #ifdef ARCH_ALPHA
51 static int coeff[64]={ 51 #define FAST_64BIT
52 W4, W4, W4, W4, W4, W4, W4, W4, 52 #endif
53 W1, W3, W5, W7,-W7,-W5,-W3,-W1, 53
54 W2, W6,-W6,-W2,-W2,-W6, W6, W2, 54 #if defined(ARCH_POWERPC_405)
55 W3,-W7,-W1,-W5, W5, W1, W7,-W3, 55
56 W4,-W4,-W4, W4, W4,-W4,-W4, W4, 56 /* signed 16x16 -> 32 multiply add accumulate */
57 W5,-W1, W7, W3,-W3,-W7, W1,-W5, 57 #define MAC16(rt, ra, rb) \
58 W6,-W2, W2,-W6,-W6, W2,-W2, W6, 58 asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb));
59 W7,-W5, W3,-W1, W1,-W3, W5,-W7 59
60 }; 60 /* signed 16x16 -> 32 multiply */
61 61 #define MUL16(rt, ra, rb) \
62 static inline int idctRowCondZ (int16_t * row) 62 asm ("mullhw %0, %1, %2" : "=r" (rt) : "r" (ra), "r" (rb));
63 { 63
64 int a0, a1, a2, a3, b0, b1, b2, b3; 64 #else
65 65
66 if( !( ((uint32_t*)row)[0]|((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3])) { 66 /* signed 16x16 -> 32 multiply add accumulate */
67 /* row[0] = row[1] = row[2] = row[3] = row[4] = 67 #define MAC16(rt, ra, rb) rt += (ra) * (rb)
68 row[5] = row[6] = row[7] = 0;*/ 68
69 return 0; 69 /* signed 16x16 -> 32 multiply */
70 } 70 #define MUL16(rt, ra, rb) rt = (ra) * (rb)
71 71
72 if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){ 72 #endif
73 a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1));
74 a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1));
75 a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1));
76 a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1));
77
78 b0 = W1*row[1] + W3*row[3];
79 b1 = W3*row[1] - W7*row[3];
80 b2 = W5*row[1] - W1*row[3];
81 b3 = W7*row[1] - W5*row[3];
82 }else{
83 a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1));
84 a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1));
85 a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1));
86 a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1));
87
88 b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7];
89 b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7];
90 b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7];
91 b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7];
92 }
93
94 row[0] = (a0 + b0) >> ROW_SHIFT;
95 row[1] = (a1 + b1) >> ROW_SHIFT;
96 row[2] = (a2 + b2) >> ROW_SHIFT;
97 row[3] = (a3 + b3) >> ROW_SHIFT;
98 row[4] = (a3 - b3) >> ROW_SHIFT;
99 row[5] = (a2 - b2) >> ROW_SHIFT;
100 row[6] = (a1 - b1) >> ROW_SHIFT;
101 row[7] = (a0 - b0) >> ROW_SHIFT;
102
103 return 1;
104 }
105 73
106 #ifdef ARCH_ALPHA 74 #ifdef ARCH_ALPHA
107 /* 0: all entries 0, 1: only first entry nonzero, 2: otherwise */ 75 /* 0: all entries 0, 1: only first entry nonzero, 2: otherwise */
108 static inline int idctRowCondDC(int16_t *row) 76 static inline int idctRowCondDC(int16_t *row)
109 { 77 {
127 95
128 return 1; 96 return 1;
129 } 97 }
130 } 98 }
131 99
132 a0 = W4 * row[0]; 100 a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
133 a1 = W4 * row[0]; 101 a1 = a0;
134 a2 = W4 * row[0]; 102 a2 = a0;
135 a3 = W4 * row[0]; 103 a3 = a0;
136 104
137 if (row[2]) { 105 if (row[2]) {
138 a0 += W2 * row[2]; 106 a0 += W2 * row[2];
139 a1 += W6 * row[2]; 107 a1 += W6 * row[2];
140 a2 -= W6 * row[2]; 108 a2 -= W6 * row[2];
152 a0 += W6 * row[6]; 120 a0 += W6 * row[6];
153 a1 -= W2 * row[6]; 121 a1 -= W2 * row[6];
154 a2 += W2 * row[6]; 122 a2 += W2 * row[6];
155 a3 -= W6 * row[6]; 123 a3 -= W6 * row[6];
156 } 124 }
157
158 a0 += 1 << (ROW_SHIFT - 1);
159 a1 += 1 << (ROW_SHIFT - 1);
160 a2 += 1 << (ROW_SHIFT - 1);
161 a3 += 1 << (ROW_SHIFT - 1);
162 125
163 if (row[1]) { 126 if (row[1]) {
164 b0 = W1 * row[1]; 127 b0 = W1 * row[1];
165 b1 = W3 * row[1]; 128 b1 = W3 * row[1];
166 b2 = W5 * row[1]; 129 b2 = W5 * row[1];
203 row[7] = (a0 - b0) >> ROW_SHIFT; 166 row[7] = (a0 - b0) >> ROW_SHIFT;
204 167
205 return 2; 168 return 2;
206 } 169 }
207 #else /* not ARCH_ALPHA */ 170 #else /* not ARCH_ALPHA */
208 static inline int idctRowCondDC (int16_t * row) 171
172 static inline void idctRowCondDC (int16_t * row)
209 { 173 {
210 int a0, a1, a2, a3, b0, b1, b2, b3; 174 int a0, a1, a2, a3, b0, b1, b2, b3;
211 175 #ifdef FAST_64BIT
212 if( !( ((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3]| row[1])) { 176 uint64_t temp;
213 // row[0] = row[1] = row[2] = row[3] = row[4] = row[5] = row[6] = row[7] = row[0]<<3; 177 #else
214 uint16_t temp= row[0]<<3; 178 uint32_t temp;
215 ((uint32_t*)row)[0]=((uint32_t*)row)[1]= 179 #endif
216 ((uint32_t*)row)[2]=((uint32_t*)row)[3]= temp + (temp<<16); 180
217 return 0; 181 #ifdef FAST_64BIT
218 } 182 #ifdef WORDS_BIGENDIAN
219 183 #define ROW0_MASK 0xffff000000000000LL
220 if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){ 184 #else
221 a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1)); 185 #define ROW0_MASK 0xffffLL
222 a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1)); 186 #endif
223 a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1)); 187 if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) |
224 a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1)); 188 ((uint64_t *)row)[1]) == 0) {
225 189 temp = (row[0] << 3) & 0xffff;
226 b0 = W1*row[1] + W3*row[3]; 190 temp += temp << 16;
227 b1 = W3*row[1] - W7*row[3]; 191 temp += temp << 32;
228 b2 = W5*row[1] - W1*row[3]; 192 ((uint64_t *)row)[0] = temp;
229 b3 = W7*row[1] - W5*row[3]; 193 ((uint64_t *)row)[1] = temp;
230 }else{ 194 return;
231 a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1)); 195 }
232 a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1)); 196 #else
233 a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1)); 197 if (!(((uint32_t*)row)[1] |
234 a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1)); 198 ((uint32_t*)row)[2] |
235 199 ((uint32_t*)row)[3] |
236 b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7]; 200 row[1])) {
237 b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7]; 201 temp = (row[0] << 3) & 0xffff;
238 b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7]; 202 temp += temp << 16;
239 b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7]; 203 ((uint32_t*)row)[0]=((uint32_t*)row)[1] =
204 ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp;
205 return;
206 }
207 #endif
208
209 a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
210 a1 = a0;
211 a2 = a0;
212 a3 = a0;
213
214 /* no need to optimize : gcc does it */
215 a0 += W2 * row[2];
216 a1 += W6 * row[2];
217 a2 -= W6 * row[2];
218 a3 -= W2 * row[2];
219
220 MUL16(b0, W1, row[1]);
221 MAC16(b0, W3, row[3]);
222 MUL16(b1, W3, row[1]);
223 MAC16(b1, -W7, row[3]);
224 MUL16(b2, W5, row[1]);
225 MAC16(b2, -W1, row[3]);
226 MUL16(b3, W7, row[1]);
227 MAC16(b3, -W5, row[3]);
228
229 #ifdef FAST_64BIT
230 temp = ((uint64_t*)row)[1];
231 #else
232 temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3];
233 #endif
234 if (temp != 0) {
235 a0 += W4*row[4] + W6*row[6];
236 a1 += - W4*row[4] - W2*row[6];
237 a2 += - W4*row[4] + W2*row[6];
238 a3 += W4*row[4] - W6*row[6];
239
240 MAC16(b0, W5, row[5]);
241 MAC16(b0, W7, row[7]);
242
243 MAC16(b1, -W1, row[5]);
244 MAC16(b1, -W5, row[7]);
245
246 MAC16(b2, W7, row[5]);
247 MAC16(b2, W3, row[7]);
248
249 MAC16(b3, W3, row[5]);
250 MAC16(b3, -W1, row[7]);
240 } 251 }
241 252
242 row[0] = (a0 + b0) >> ROW_SHIFT; 253 row[0] = (a0 + b0) >> ROW_SHIFT;
243 row[7] = (a0 - b0) >> ROW_SHIFT; 254 row[7] = (a0 - b0) >> ROW_SHIFT;
244 row[1] = (a1 + b1) >> ROW_SHIFT; 255 row[1] = (a1 + b1) >> ROW_SHIFT;
245 row[6] = (a1 - b1) >> ROW_SHIFT; 256 row[6] = (a1 - b1) >> ROW_SHIFT;
246 row[2] = (a2 + b2) >> ROW_SHIFT; 257 row[2] = (a2 + b2) >> ROW_SHIFT;
247 row[5] = (a2 - b2) >> ROW_SHIFT; 258 row[5] = (a2 - b2) >> ROW_SHIFT;
248 row[3] = (a3 + b3) >> ROW_SHIFT; 259 row[3] = (a3 + b3) >> ROW_SHIFT;
249 row[4] = (a3 - b3) >> ROW_SHIFT; 260 row[4] = (a3 - b3) >> ROW_SHIFT;
250
251 return 1;
252 } 261 }
253 #endif /* not ARCH_ALPHA */ 262 #endif /* not ARCH_ALPHA */
254 263
255 static inline void idctCol (int16_t * col) 264 static inline void idctSparseCol (int16_t * col)
256 { 265 {
257
258 /*
259 if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) {
260 col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] =
261 col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3;
262 return;
263 }*/
264
265 int a0, a1, a2, a3, b0, b1, b2, b3; 266 int a0, a1, a2, a3, b0, b1, b2, b3;
266 col[0] += (1<<(COL_SHIFT-1))/W4; 267
267 a0 = W4*col[8*0] + W2*col[8*2] + W4*col[8*4] + W6*col[8*6]; 268 /* XXX: I did that only to give same values as previous code */
268 a1 = W4*col[8*0] + W6*col[8*2] - W4*col[8*4] - W2*col[8*6]; 269 a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
269 a2 = W4*col[8*0] - W6*col[8*2] - W4*col[8*4] + W2*col[8*6]; 270 a1 = a0;
270 a3 = W4*col[8*0] - W2*col[8*2] + W4*col[8*4] - W6*col[8*6]; 271 a2 = a0;
271 272 a3 = a0;
272 b0 = W1*col[8*1] + W3*col[8*3] + W5*col[8*5] + W7*col[8*7]; 273
273 b1 = W3*col[8*1] - W7*col[8*3] - W1*col[8*5] - W5*col[8*7]; 274 a0 += + W2*col[8*2];
274 b2 = W5*col[8*1] - W1*col[8*3] + W7*col[8*5] + W3*col[8*7]; 275 a1 += + W6*col[8*2];
275 b3 = W7*col[8*1] - W5*col[8*3] + W3*col[8*5] - W1*col[8*7]; 276 a2 += - W6*col[8*2];
276 277 a3 += - W2*col[8*2];
277 col[8*0] = (a0 + b0) >> COL_SHIFT; 278
278 col[8*7] = (a0 - b0) >> COL_SHIFT; 279 MUL16(b0, W1, col[8*1]);
279 col[8*1] = (a1 + b1) >> COL_SHIFT; 280 MUL16(b1, W3, col[8*1]);
280 col[8*6] = (a1 - b1) >> COL_SHIFT; 281 MUL16(b2, W5, col[8*1]);
281 col[8*2] = (a2 + b2) >> COL_SHIFT; 282 MUL16(b3, W7, col[8*1]);
282 col[8*5] = (a2 - b2) >> COL_SHIFT; 283
283 col[8*3] = (a3 + b3) >> COL_SHIFT; 284 MAC16(b0, + W3, col[8*3]);
284 col[8*4] = (a3 - b3) >> COL_SHIFT; 285 MAC16(b1, - W7, col[8*3]);
285 } 286 MAC16(b2, - W1, col[8*3]);
286 287 MAC16(b3, - W5, col[8*3]);
287 static inline void idctSparseCol (int16_t * col)
288 {
289 int a0, a1, a2, a3, b0, b1, b2, b3;
290 col[0] += (1<<(COL_SHIFT-1))/W4;
291 a0 = W4*col[8*0];
292 a1 = W4*col[8*0];
293 a2 = W4*col[8*0];
294 a3 = W4*col[8*0];
295
296 if(col[8*2]){
297 a0 += + W2*col[8*2];
298 a1 += + W6*col[8*2];
299 a2 += - W6*col[8*2];
300 a3 += - W2*col[8*2];
301 }
302 288
303 if(col[8*4]){ 289 if(col[8*4]){
304 a0 += + W4*col[8*4]; 290 a0 += + W4*col[8*4];
305 a1 += - W4*col[8*4]; 291 a1 += - W4*col[8*4];
306 a2 += - W4*col[8*4]; 292 a2 += - W4*col[8*4];
307 a3 += + W4*col[8*4]; 293 a3 += + W4*col[8*4];
294 }
295
296 if (col[8*5]) {
297 MAC16(b0, + W5, col[8*5]);
298 MAC16(b1, - W1, col[8*5]);
299 MAC16(b2, + W7, col[8*5]);
300 MAC16(b3, + W3, col[8*5]);
308 } 301 }
309 302
310 if(col[8*6]){ 303 if(col[8*6]){
311 a0 += + W6*col[8*6]; 304 a0 += + W6*col[8*6];
312 a1 += - W2*col[8*6]; 305 a1 += - W2*col[8*6];
313 a2 += + W2*col[8*6]; 306 a2 += + W2*col[8*6];
314 a3 += - W6*col[8*6]; 307 a3 += - W6*col[8*6];
315 } 308 }
316 309
317 if(col[8*1]){ 310 if (col[8*7]) {
318 b0 = W1*col[8*1]; 311 MAC16(b0, + W7, col[8*7]);
319 b1 = W3*col[8*1]; 312 MAC16(b1, - W5, col[8*7]);
320 b2 = W5*col[8*1]; 313 MAC16(b2, + W3, col[8*7]);
321 b3 = W7*col[8*1]; 314 MAC16(b3, - W1, col[8*7]);
322 }else{ 315 }
323 b0 = 316
324 b1 = 317 col[8*0] = (a0 + b0) >> COL_SHIFT;
325 b2 = 318 col[8*7] = (a0 - b0) >> COL_SHIFT;
326 b3 = 0; 319 col[8*1] = (a1 + b1) >> COL_SHIFT;
327 } 320 col[8*6] = (a1 - b1) >> COL_SHIFT;
328 321 col[8*2] = (a2 + b2) >> COL_SHIFT;
329 if(col[8*3]){ 322 col[8*5] = (a2 - b2) >> COL_SHIFT;
330 b0 += + W3*col[8*3]; 323 col[8*3] = (a3 + b3) >> COL_SHIFT;
331 b1 += - W7*col[8*3]; 324 col[8*4] = (a3 - b3) >> COL_SHIFT;
332 b2 += - W1*col[8*3];
333 b3 += - W5*col[8*3];
334 }
335
336 if(col[8*5]){
337 b0 += + W5*col[8*5];
338 b1 += - W1*col[8*5];
339 b2 += + W7*col[8*5];
340 b3 += + W3*col[8*5];
341 }
342
343 if(col[8*7]){
344 b0 += + W7*col[8*7];
345 b1 += - W5*col[8*7];
346 b2 += + W3*col[8*7];
347 b3 += - W1*col[8*7];
348 }
349
350 #ifndef ARCH_ALPHA
351 if(!(b0|b1|b2|b3)){
352 col[8*0] = (a0) >> COL_SHIFT;
353 col[8*7] = (a0) >> COL_SHIFT;
354 col[8*1] = (a1) >> COL_SHIFT;
355 col[8*6] = (a1) >> COL_SHIFT;
356 col[8*2] = (a2) >> COL_SHIFT;
357 col[8*5] = (a2) >> COL_SHIFT;
358 col[8*3] = (a3) >> COL_SHIFT;
359 col[8*4] = (a3) >> COL_SHIFT;
360 }else{
361 #endif
362 col[8*0] = (a0 + b0) >> COL_SHIFT;
363 col[8*7] = (a0 - b0) >> COL_SHIFT;
364 col[8*1] = (a1 + b1) >> COL_SHIFT;
365 col[8*6] = (a1 - b1) >> COL_SHIFT;
366 col[8*2] = (a2 + b2) >> COL_SHIFT;
367 col[8*5] = (a2 - b2) >> COL_SHIFT;
368 col[8*3] = (a3 + b3) >> COL_SHIFT;
369 col[8*4] = (a3 - b3) >> COL_SHIFT;
370 #ifndef ARCH_ALPHA
371 }
372 #endif
373 }
374
375 static inline void idctSparse2Col (int16_t * col)
376 {
377 int a0, a1, a2, a3, b0, b1, b2, b3;
378 col[0] += (1<<(COL_SHIFT-1))/W4;
379 a0 = W4*col[8*0];
380 a1 = W4*col[8*0];
381 a2 = W4*col[8*0];
382 a3 = W4*col[8*0];
383
384 if(col[8*2]){
385 a0 += + W2*col[8*2];
386 a1 += + W6*col[8*2];
387 a2 += - W6*col[8*2];
388 a3 += - W2*col[8*2];
389 }
390
391 if(col[8*4]){
392 a0 += + W4*col[8*4];
393 a1 += - W4*col[8*4];
394 a2 += - W4*col[8*4];
395 a3 += + W4*col[8*4];
396 }
397
398 if(col[8*6]){
399 a0 += + W6*col[8*6];
400 a1 += - W2*col[8*6];
401 a2 += + W2*col[8*6];
402 a3 += - W6*col[8*6];
403 }
404
405 if(col[8*1] || 1){
406 b0 = W1*col[8*1];
407 b1 = W3*col[8*1];
408 b2 = W5*col[8*1];
409 b3 = W7*col[8*1];
410 }else{
411 b0 =
412 b1 =
413 b2 =
414 b3 = 0;
415 }
416
417 if(col[8*3]){
418 b0 += + W3*col[8*3];
419 b1 += - W7*col[8*3];
420 b2 += - W1*col[8*3];
421 b3 += - W5*col[8*3];
422 }
423
424 if(col[8*5]){
425 b0 += + W5*col[8*5];
426 b1 += - W1*col[8*5];
427 b2 += + W7*col[8*5];
428 b3 += + W3*col[8*5];
429 }
430
431 if(col[8*7]){
432 b0 += + W7*col[8*7];
433 b1 += - W5*col[8*7];
434 b2 += + W3*col[8*7];
435 b3 += - W1*col[8*7];
436 }
437
438 col[8*0] = (a0 + b0) >> COL_SHIFT;
439 col[8*7] = (a0 - b0) >> COL_SHIFT;
440 col[8*1] = (a1 + b1) >> COL_SHIFT;
441 col[8*6] = (a1 - b1) >> COL_SHIFT;
442 col[8*2] = (a2 + b2) >> COL_SHIFT;
443 col[8*5] = (a2 - b2) >> COL_SHIFT;
444 col[8*3] = (a3 + b3) >> COL_SHIFT;
445 col[8*4] = (a3 - b3) >> COL_SHIFT;
446 } 325 }
447 326
448 #ifdef ARCH_ALPHA 327 #ifdef ARCH_ALPHA
449 /* If all rows but the first one are zero after row transformation, 328 /* If all rows but the first one are zero after row transformation,
450 all rows will be identical after column transformation. */ 329 all rows will be identical after column transformation. */
470 lcol[ 8] = l; lcol[ 9] = r; 349 lcol[ 8] = l; lcol[ 9] = r;
471 lcol[10] = l; lcol[11] = r; 350 lcol[10] = l; lcol[11] = r;
472 lcol[12] = l; lcol[13] = r; 351 lcol[12] = l; lcol[13] = r;
473 lcol[14] = l; lcol[15] = r; 352 lcol[14] = l; lcol[15] = r;
474 } 353 }
475 #endif
476 354
477 void simple_idct (short *block) 355 void simple_idct (short *block)
478 { 356 {
479 357
480 int i; 358 int i;
481
482 #if 0
483 int nonZero[8];
484 int buffer[64];
485 int nNonZero=0;
486
487 idctRowCondDC(block);
488
489 for(i=1; i<8; i++)
490 {
491 nonZero[nNonZero]=i;
492 nNonZero+= idctRowCondZ(block + i*8);
493 }
494
495 if(nNonZero==0)
496 {
497 for(i=0; i<8; i++)
498 {
499 block[i ]=
500 block[i+8 ]=
501 block[i+16]=
502 block[i+24]=
503 block[i+32]=
504 block[i+40]=
505 block[i+48]=
506 block[i+56]= (W4*block[i] + (1<<(COL_SHIFT-1))) >> COL_SHIFT;
507 }
508 }
509 else if(nNonZero==1)
510 {
511 int index= nonZero[0]*8;
512 for(i=0; i<8; i++)
513 {
514 int bias= W4*block[i] + (1<<(COL_SHIFT-1));
515 int c= block[i + index];
516 block[i ]= (c*coeff[index ] + bias) >> COL_SHIFT;
517 block[i+8 ]= (c*coeff[index+1] + bias) >> COL_SHIFT;
518 block[i+16]= (c*coeff[index+2] + bias) >> COL_SHIFT;
519 block[i+24]= (c*coeff[index+3] + bias) >> COL_SHIFT;
520 block[i+32]= (c*coeff[index+4] + bias) >> COL_SHIFT;
521 block[i+40]= (c*coeff[index+5] + bias) >> COL_SHIFT;
522 block[i+48]= (c*coeff[index+6] + bias) >> COL_SHIFT;
523 block[i+56]= (c*coeff[index+7] + bias) >> COL_SHIFT;
524 }
525 }
526 /* else if(nNonZero==2)
527 {
528 int index1= nonZero[0]*8;
529 int index2= nonZero[1]*8;
530 for(i=0; i<8; i++)
531 {
532 int bias= W4*block[i] + (1<<(COL_SHIFT-1));
533 int c1= block[i + index1];
534 int c2= block[i + index2];
535 block[i ]= (c1*coeff[index1 ] + c2*coeff[index2 ] + bias) >> COL_SHIFT;
536 block[i+8 ]= (c1*coeff[index1+1] + c2*coeff[index2+1] + bias) >> COL_SHIFT;
537 block[i+16]= (c1*coeff[index1+2] + c2*coeff[index2+2] + bias) >> COL_SHIFT;
538 block[i+24]= (c1*coeff[index1+3] + c2*coeff[index2+3] + bias) >> COL_SHIFT;
539 block[i+32]= (c1*coeff[index1+4] + c2*coeff[index2+4] + bias) >> COL_SHIFT;
540 block[i+40]= (c1*coeff[index1+5] + c2*coeff[index2+5] + bias) >> COL_SHIFT;
541 block[i+48]= (c1*coeff[index1+6] + c2*coeff[index2+6] + bias) >> COL_SHIFT;
542 block[i+56]= (c1*coeff[index1+7] + c2*coeff[index2+7] + bias) >> COL_SHIFT;
543 }
544 }*/
545 else
546 {
547 for(i=0; i<8; i++)
548 idctSparse2Col(block + i);
549 }
550 #elif defined(ARCH_ALPHA)
551 int rowsZero = 1; /* all rows except row 0 zero */ 359 int rowsZero = 1; /* all rows except row 0 zero */
552 int rowsConstant = 1; /* all rows consist of a constant value */ 360 int rowsConstant = 1; /* all rows consist of a constant value */
553 361
554 for (i = 0; i < 8; i++) { 362 for (i = 0; i < 8; i++) {
555 int sparseness = idctRowCondDC(block + 8 * i); 363 int sparseness = idctRowCondDC(block + 8 * i);
577 } 385 }
578 } else { 386 } else {
579 for (i = 0; i < 8; i++) 387 for (i = 0; i < 8; i++)
580 idctSparseCol(block + i); 388 idctSparseCol(block + i);
581 } 389 }
582 #else 390 }
583 for(i=0; i<8; i++) 391
584 idctRowCondDC(block + i*8); 392 #else
585 393
586 for(i=0; i<8; i++) 394 void simple_idct (short *block)
587 idctSparseCol(block + i); 395 {
588 #endif 396 int i;
589 } 397 for(i=0; i<8; i++)
398 idctRowCondDC(block + i*8);
399
400 for(i=0; i<8; i++)
401 idctSparseCol(block + i);
402 }
403
404 #endif
590 405
591 #undef COL_SHIFT 406 #undef COL_SHIFT