Mercurial > libavcodec.hg

comparison simple_idct.c @ 205:ccf36af385f3 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
(commit by michael) faster simple_idct in C
author arpi_esp
date Mon, 14 Jan 2002 04:39:59 +0000
parents bd77d3cbb233
children 73df666cacc7
comparison
equal deleted inserted replaced
204:fceb435fae6b 205:ccf36af385f3
43 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 43 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
44 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 44 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
45 #define ROW_SHIFT 11 45 #define ROW_SHIFT 11
46 #define COL_SHIFT 20 // 6 46 #define COL_SHIFT 20 // 6
47 #endif 47 #endif
48 #if 1 48
49 static void inline idctRow (int16_t * row) 49 /* 8x8 Matrix used to do a trivial (slow) 8 point IDCT */
50 static int coeff[64]={
51 W4, W4, W4, W4, W4, W4, W4, W4,
52 W1, W3, W5, W7,-W7,-W5,-W3,-W1,
53 W2, W6,-W6,-W2,-W2,-W6, W6, W2,
54 W3,-W7,-W1,-W5, W5, W1, W7,-W3,
55 W4,-W4,-W4, W4, W4,-W4,-W4, W4,
56 W5,-W1, W7, W3,-W3,-W7, W1,-W5,
57 W6,-W2, W2,-W6,-W6, W2,-W2, W6,
58 W7,-W5, W3,-W1, W1,-W3, W5,-W7
59 };
60
61 static int inline idctRowCondZ (int16_t * row)
50 { 62 {
51 int a0, a1, a2, a3, b0, b1, b2, b3; 63 int a0, a1, a2, a3, b0, b1, b2, b3;
52 const int C1 =W1; 64
53 const int C2 =W2; 65 if( !( ((uint32_t*)row)[0]|((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3])) {
54 const int C3 =W3; 66 /* row[0] = row[1] = row[2] = row[3] = row[4] =
55 const int C4 =W4; 67 row[5] = row[6] = row[7] = 0;*/
56 const int C5 =W5; 68 return 0;
57 const int C6 =W6; 69 }
58 const int C7 =W7; 70
59 71 if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){
60 if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7])) { 72 a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1));
61 row[0] = row[1] = row[2] = row[3] = row[4] = 73 a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1));
62 row[5] = row[6] = row[7] = row[0]<<3; 74 a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1));
63 return; 75 a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1));
64 } 76
65 77 b0 = W1*row[1] + W3*row[3];
66 a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + (1<<(ROW_SHIFT-1)); 78 b1 = W3*row[1] - W7*row[3];
67 a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + (1<<(ROW_SHIFT-1)); 79 b2 = W5*row[1] - W1*row[3];
68 a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + (1<<(ROW_SHIFT-1)); 80 b3 = W7*row[1] - W5*row[3];
69 a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + (1<<(ROW_SHIFT-1)); 81 }else{
70 82 a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1));
71 b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7]; 83 a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1));
72 b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7]; 84 a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1));
73 b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7]; 85 a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1));
74 b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7]; 86
87 b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7];
88 b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7];
89 b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7];
90 b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7];
91 }
75 92
76 row[0] = (a0 + b0) >> ROW_SHIFT; 93 row[0] = (a0 + b0) >> ROW_SHIFT;
77 row[1] = (a1 + b1) >> ROW_SHIFT; 94 row[1] = (a1 + b1) >> ROW_SHIFT;
78 row[2] = (a2 + b2) >> ROW_SHIFT; 95 row[2] = (a2 + b2) >> ROW_SHIFT;
79 row[3] = (a3 + b3) >> ROW_SHIFT; 96 row[3] = (a3 + b3) >> ROW_SHIFT;
80 row[4] = (a3 - b3) >> ROW_SHIFT; 97 row[4] = (a3 - b3) >> ROW_SHIFT;
81 row[5] = (a2 - b2) >> ROW_SHIFT; 98 row[5] = (a2 - b2) >> ROW_SHIFT;
82 row[6] = (a1 - b1) >> ROW_SHIFT; 99 row[6] = (a1 - b1) >> ROW_SHIFT;
83 row[7] = (a0 - b0) >> ROW_SHIFT; 100 row[7] = (a0 - b0) >> ROW_SHIFT;
101
102 return 1;
103 }
104
105 static int inline idctRowCondDC (int16_t * row)
106 {
107 int a0, a1, a2, a3, b0, b1, b2, b3;
108
109 if( !( ((uint32_t*)row)[1] |((uint32_t*)row)[2] |((uint32_t*)row)[3]| row[1])) {
110 // row[0] = row[1] = row[2] = row[3] = row[4] = row[5] = row[6] = row[7] = row[0]<<3;
111 uint16_t temp= row[0]<<3;
112 ((uint32_t*)row)[0]=((uint32_t*)row)[1]=
113 ((uint32_t*)row)[2]=((uint32_t*)row)[3]= temp + (temp<<16);
114 return 0;
115 }
116
117 if(!( ((uint32_t*)row)[2] |((uint32_t*)row)[3] )){
118 a0 = W4*row[0] + W2*row[2] + (1<<(ROW_SHIFT-1));
119 a1 = W4*row[0] + W6*row[2] + (1<<(ROW_SHIFT-1));
120 a2 = W4*row[0] - W6*row[2] + (1<<(ROW_SHIFT-1));
121 a3 = W4*row[0] - W2*row[2] + (1<<(ROW_SHIFT-1));
122
123 b0 = W1*row[1] + W3*row[3];
124 b1 = W3*row[1] - W7*row[3];
125 b2 = W5*row[1] - W1*row[3];
126 b3 = W7*row[1] - W5*row[3];
127 }else{
128 a0 = W4*row[0] + W2*row[2] + W4*row[4] + W6*row[6] + (1<<(ROW_SHIFT-1));
129 a1 = W4*row[0] + W6*row[2] - W4*row[4] - W2*row[6] + (1<<(ROW_SHIFT-1));
130 a2 = W4*row[0] - W6*row[2] - W4*row[4] + W2*row[6] + (1<<(ROW_SHIFT-1));
131 a3 = W4*row[0] - W2*row[2] + W4*row[4] - W6*row[6] + (1<<(ROW_SHIFT-1));
132
133 b0 = W1*row[1] + W3*row[3] + W5*row[5] + W7*row[7];
134 b1 = W3*row[1] - W7*row[3] - W1*row[5] - W5*row[7];
135 b2 = W5*row[1] - W1*row[3] + W7*row[5] + W3*row[7];
136 b3 = W7*row[1] - W5*row[3] + W3*row[5] - W1*row[7];
137 }
138
139 row[0] = (a0 + b0) >> ROW_SHIFT;
140 row[7] = (a0 - b0) >> ROW_SHIFT;
141 row[1] = (a1 + b1) >> ROW_SHIFT;
142 row[6] = (a1 - b1) >> ROW_SHIFT;
143 row[2] = (a2 + b2) >> ROW_SHIFT;
144 row[5] = (a2 - b2) >> ROW_SHIFT;
145 row[3] = (a3 + b3) >> ROW_SHIFT;
146 row[4] = (a3 - b3) >> ROW_SHIFT;
147
148 return 1;
84 } 149 }
85 150
86 static void inline idctCol (int16_t * col) 151 static void inline idctCol (int16_t * col)
87 { 152 {
88 int a0, a1, a2, a3, b0, b1, b2, b3; 153
89 const int C1 =W1;
90 const int C2 =W2;
91 const int C3 =W3;
92 const int C4 =W4;
93 const int C5 =W5;
94 const int C6 =W6;
95 const int C7 =W7;
96 /* 154 /*
97 if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) { 155 if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) {
98 col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] = 156 col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] =
99 col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3; 157 col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3;
100 return; 158 return;
101 }*/ 159 }*/
160
161 int a0, a1, a2, a3, b0, b1, b2, b3;
102 col[0] += (1<<(COL_SHIFT-1))/W4; 162 col[0] += (1<<(COL_SHIFT-1))/W4;
103 a0 = C4*col[8*0] + C2*col[8*2] + C4*col[8*4] + C6*col[8*6]; 163 a0 = W4*col[8*0] + W2*col[8*2] + W4*col[8*4] + W6*col[8*6];
104 a1 = C4*col[8*0] + C6*col[8*2] - C4*col[8*4] - C2*col[8*6]; 164 a1 = W4*col[8*0] + W6*col[8*2] - W4*col[8*4] - W2*col[8*6];
105 a2 = C4*col[8*0] - C6*col[8*2] - C4*col[8*4] + C2*col[8*6]; 165 a2 = W4*col[8*0] - W6*col[8*2] - W4*col[8*4] + W2*col[8*6];
106 a3 = C4*col[8*0] - C2*col[8*2] + C4*col[8*4] - C6*col[8*6]; 166 a3 = W4*col[8*0] - W2*col[8*2] + W4*col[8*4] - W6*col[8*6];
107 167
108 b0 = C1*col[8*1] + C3*col[8*3] + C5*col[8*5] + C7*col[8*7]; 168 b0 = W1*col[8*1] + W3*col[8*3] + W5*col[8*5] + W7*col[8*7];
109 b1 = C3*col[8*1] - C7*col[8*3] - C1*col[8*5] - C5*col[8*7]; 169 b1 = W3*col[8*1] - W7*col[8*3] - W1*col[8*5] - W5*col[8*7];
110 b2 = C5*col[8*1] - C1*col[8*3] + C7*col[8*5] + C3*col[8*7]; 170 b2 = W5*col[8*1] - W1*col[8*3] + W7*col[8*5] + W3*col[8*7];
111 b3 = C7*col[8*1] - C5*col[8*3] + C3*col[8*5] - C1*col[8*7]; 171 b3 = W7*col[8*1] - W5*col[8*3] + W3*col[8*5] - W1*col[8*7];
112 172
113 col[8*0] = (a0 + b0) >> COL_SHIFT; 173 col[8*0] = (a0 + b0) >> COL_SHIFT;
174 col[8*7] = (a0 - b0) >> COL_SHIFT;
114 col[8*1] = (a1 + b1) >> COL_SHIFT; 175 col[8*1] = (a1 + b1) >> COL_SHIFT;
176 col[8*6] = (a1 - b1) >> COL_SHIFT;
115 col[8*2] = (a2 + b2) >> COL_SHIFT; 177 col[8*2] = (a2 + b2) >> COL_SHIFT;
178 col[8*5] = (a2 - b2) >> COL_SHIFT;
116 col[8*3] = (a3 + b3) >> COL_SHIFT; 179 col[8*3] = (a3 + b3) >> COL_SHIFT;
117 col[8*4] = (a3 - b3) >> COL_SHIFT; 180 col[8*4] = (a3 - b3) >> COL_SHIFT;
181 }
182
183 static void inline idctSparseCol (int16_t * col)
184 {
185 int a0, a1, a2, a3, b0, b1, b2, b3;
186 col[0] += (1<<(COL_SHIFT-1))/W4;
187 a0 = W4*col[8*0];
188 a1 = W4*col[8*0];
189 a2 = W4*col[8*0];
190 a3 = W4*col[8*0];
191
192 if(col[8*2]){
193 a0 += + W2*col[8*2];
194 a1 += + W6*col[8*2];
195 a2 += - W6*col[8*2];
196 a3 += - W2*col[8*2];
197 }
198
199 if(col[8*4]){
200 a0 += + W4*col[8*4];
201 a1 += - W4*col[8*4];
202 a2 += - W4*col[8*4];
203 a3 += + W4*col[8*4];
204 }
205
206 if(col[8*6]){
207 a0 += + W6*col[8*6];
208 a1 += - W2*col[8*6];
209 a2 += + W2*col[8*6];
210 a3 += - W6*col[8*6];
211 }
212
213 if(col[8*1]){
214 b0 = W1*col[8*1];
215 b1 = W3*col[8*1];
216 b2 = W5*col[8*1];
217 b3 = W7*col[8*1];
218 }else{
219 b0 =
220 b1 =
221 b2 =
222 b3 = 0;
223 }
224
225 if(col[8*3]){
226 b0 += + W3*col[8*3];
227 b1 += - W7*col[8*3];
228 b2 += - W1*col[8*3];
229 b3 += - W5*col[8*3];
230 }
231
232 if(col[8*5]){
233 b0 += + W5*col[8*5];
234 b1 += - W1*col[8*5];
235 b2 += + W7*col[8*5];
236 b3 += + W3*col[8*5];
237 }
238
239 if(col[8*7]){
240 b0 += + W7*col[8*7];
241 b1 += - W5*col[8*7];
242 b2 += + W3*col[8*7];
243 b3 += - W1*col[8*7];
244 }
245
246 if(!(b0|b1|b2|b3)){
247 col[8*0] = (a0) >> COL_SHIFT;
248 col[8*7] = (a0) >> COL_SHIFT;
249 col[8*1] = (a1) >> COL_SHIFT;
250 col[8*6] = (a1) >> COL_SHIFT;
251 col[8*2] = (a2) >> COL_SHIFT;
252 col[8*5] = (a2) >> COL_SHIFT;
253 col[8*3] = (a3) >> COL_SHIFT;
254 col[8*4] = (a3) >> COL_SHIFT;
255 }else{
256 col[8*0] = (a0 + b0) >> COL_SHIFT;
257 col[8*7] = (a0 - b0) >> COL_SHIFT;
258 col[8*1] = (a1 + b1) >> COL_SHIFT;
259 col[8*6] = (a1 - b1) >> COL_SHIFT;
260 col[8*2] = (a2 + b2) >> COL_SHIFT;
261 col[8*5] = (a2 - b2) >> COL_SHIFT;
262 col[8*3] = (a3 + b3) >> COL_SHIFT;
263 col[8*4] = (a3 - b3) >> COL_SHIFT;
264 }
265 }
266
267 static void inline idctSparse2Col (int16_t * col)
268 {
269 int a0, a1, a2, a3, b0, b1, b2, b3;
270 col[0] += (1<<(COL_SHIFT-1))/W4;
271 a0 = W4*col[8*0];
272 a1 = W4*col[8*0];
273 a2 = W4*col[8*0];
274 a3 = W4*col[8*0];
275
276 if(col[8*2]){
277 a0 += + W2*col[8*2];
278 a1 += + W6*col[8*2];
279 a2 += - W6*col[8*2];
280 a3 += - W2*col[8*2];
281 }
282
283 if(col[8*4]){
284 a0 += + W4*col[8*4];
285 a1 += - W4*col[8*4];
286 a2 += - W4*col[8*4];
287 a3 += + W4*col[8*4];
288 }
289
290 if(col[8*6]){
291 a0 += + W6*col[8*6];
292 a1 += - W2*col[8*6];
293 a2 += + W2*col[8*6];
294 a3 += - W6*col[8*6];
295 }
296
297 if(col[8*1] || 1){
298 b0 = W1*col[8*1];
299 b1 = W3*col[8*1];
300 b2 = W5*col[8*1];
301 b3 = W7*col[8*1];
302 }else{
303 b0 =
304 b1 =
305 b2 =
306 b3 = 0;
307 }
308
309 if(col[8*3]){
310 b0 += + W3*col[8*3];
311 b1 += - W7*col[8*3];
312 b2 += - W1*col[8*3];
313 b3 += - W5*col[8*3];
314 }
315
316 if(col[8*5]){
317 b0 += + W5*col[8*5];
318 b1 += - W1*col[8*5];
319 b2 += + W7*col[8*5];
320 b3 += + W3*col[8*5];
321 }
322
323 if(col[8*7]){
324 b0 += + W7*col[8*7];
325 b1 += - W5*col[8*7];
326 b2 += + W3*col[8*7];
327 b3 += - W1*col[8*7];
328 }
329
330 col[8*0] = (a0 + b0) >> COL_SHIFT;
331 col[8*7] = (a0 - b0) >> COL_SHIFT;
332 col[8*1] = (a1 + b1) >> COL_SHIFT;
333 col[8*6] = (a1 - b1) >> COL_SHIFT;
334 col[8*2] = (a2 + b2) >> COL_SHIFT;
118 col[8*5] = (a2 - b2) >> COL_SHIFT; 335 col[8*5] = (a2 - b2) >> COL_SHIFT;
119 col[8*6] = (a1 - b1) >> COL_SHIFT;
120 col[8*7] = (a0 - b0) >> COL_SHIFT;
121 }
122
123 void simple_idct (short *block)
124 {
125 int i;
126 for(i=0; i<8; i++)
127 idctRow(block + 8*i);
128
129 for(i=0; i<8; i++)
130 idctCol(block + i);
131
132 }
133
134 #else
135
136 #define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
137 #define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
138 #define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
139 #define W4 16384 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
140 #define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
141 #define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
142 #define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
143 #define COL_SHIFT 31 // 6
144
145 static void inline idctRow (int32_t *out, int16_t * row)
146 {
147 int a0, a1, a2, a3, b0, b1, b2, b3;
148 const int C1 =W1;
149 const int C2 =W2;
150 const int C3 =W3;
151 const int C4 =W4;
152 const int C5 =W5;
153 const int C6 =W6;
154 const int C7 =W7;
155 /*
156 if( !(row[1] | row[2] |row[3] |row[4] |row[5] |row[6] | row[7])) {
157 row[0] = row[1] = row[2] = row[3] = row[4] =
158 row[5] = row[6] = row[7] = row[0]<<14;
159 return;
160 }
161 */
162 a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6];
163 a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6];
164 a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6];
165 a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6];
166
167 b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
168 b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
169 b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
170 b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7];
171
172 out[0] = (a0 + b0);
173 out[1] = (a1 + b1);
174 out[2] = (a2 + b2);
175 out[3] = (a3 + b3);
176 out[4] = (a3 - b3);
177 out[5] = (a2 - b2);
178 out[6] = (a1 - b1);
179 out[7] = (a0 - b0);
180 }
181
182 static void inline idctCol (int32_t *in, int16_t * col)
183 {
184 int64_t a0, a1, a2, a3, b0, b1, b2, b3;
185 const int64_t C1 =W1;
186 const int64_t C2 =W2;
187 const int64_t C3 =W3;
188 const int64_t C4 =W4;
189 const int64_t C5 =W5;
190 const int64_t C6 =W6;
191 const int64_t C7 =W7;
192 /*
193 if( !(col[8*1] | col[8*2] |col[8*3] |col[8*4] |col[8*5] |col[8*6] | col[8*7])) {
194 col[8*0] = col[8*1] = col[8*2] = col[8*3] = col[8*4] =
195 col[8*5] = col[8*6] = col[8*7] = col[8*0]<<3;
196 return;
197 }*/
198 in[0] += (1<<(COL_SHIFT-1))/W4;
199 a0 = C4*in[8*0] + C2*in[8*2] + C4*in[8*4] + C6*in[8*6];
200 a1 = C4*in[8*0] + C6*in[8*2] - C4*in[8*4] - C2*in[8*6];
201 a2 = C4*in[8*0] - C6*in[8*2] - C4*in[8*4] + C2*in[8*6];
202 a3 = C4*in[8*0] - C2*in[8*2] + C4*in[8*4] - C6*in[8*6];
203
204 b0 = C1*in[8*1] + C3*in[8*3] + C5*in[8*5] + C7*in[8*7];
205 b1 = C3*in[8*1] - C7*in[8*3] - C1*in[8*5] - C5*in[8*7];
206 b2 = C5*in[8*1] - C1*in[8*3] + C7*in[8*5] + C3*in[8*7];
207 b3 = C7*in[8*1] - C5*in[8*3] + C3*in[8*5] - C1*in[8*7];
208
209 col[8*0] = (a0 + b0) >> COL_SHIFT;
210 col[8*1] = (a1 + b1) >> COL_SHIFT;
211 col[8*2] = (a2 + b2) >> COL_SHIFT;
212 col[8*3] = (a3 + b3) >> COL_SHIFT; 336 col[8*3] = (a3 + b3) >> COL_SHIFT;
213 col[8*4] = (a3 - b3) >> COL_SHIFT; 337 col[8*4] = (a3 - b3) >> COL_SHIFT;
214 col[8*5] = (a2 - b2) >> COL_SHIFT; 338 }
215 col[8*6] = (a1 - b1) >> COL_SHIFT; 339
216 col[8*7] = (a0 - b0) >> COL_SHIFT;
217 }
218 340
219 void simple_idct (short *block) 341 void simple_idct (short *block)
220 { 342 {
343
221 int i; 344 int i;
222 int32_t temp[64]; 345
346 #if 0
347 int nonZero[8];
348 int buffer[64];
349 int nNonZero=0;
350
351 idctRowCondDC(block);
352
353 for(i=1; i<8; i++)
354 {
355 nonZero[nNonZero]=i;
356 nNonZero+= idctRowCondZ(block + i*8);
357 }
358
359 if(nNonZero==0)
360 {
361 for(i=0; i<8; i++)
362 {
363 block[i ]=
364 block[i+8 ]=
365 block[i+16]=
366 block[i+24]=
367 block[i+32]=
368 block[i+40]=
369 block[i+48]=
370 block[i+56]= (W4*block[i] + (1<<(COL_SHIFT-1))) >> COL_SHIFT;
371 }
372 }
373 else if(nNonZero==1)
374 {
375 int index= nonZero[0]*8;
376 for(i=0; i<8; i++)
377 {
378 int bias= W4*block[i] + (1<<(COL_SHIFT-1));
379 int c= block[i + index];
380 block[i ]= (c*coeff[index ] + bias) >> COL_SHIFT;
381 block[i+8 ]= (c*coeff[index+1] + bias) >> COL_SHIFT;
382 block[i+16]= (c*coeff[index+2] + bias) >> COL_SHIFT;
383 block[i+24]= (c*coeff[index+3] + bias) >> COL_SHIFT;
384 block[i+32]= (c*coeff[index+4] + bias) >> COL_SHIFT;
385 block[i+40]= (c*coeff[index+5] + bias) >> COL_SHIFT;
386 block[i+48]= (c*coeff[index+6] + bias) >> COL_SHIFT;
387 block[i+56]= (c*coeff[index+7] + bias) >> COL_SHIFT;
388 }
389 }
390 /* else if(nNonZero==2)
391 {
392 int index1= nonZero[0]*8;
393 int index2= nonZero[1]*8;
394 for(i=0; i<8; i++)
395 {
396 int bias= W4*block[i] + (1<<(COL_SHIFT-1));
397 int c1= block[i + index1];
398 int c2= block[i + index2];
399 block[i ]= (c1*coeff[index1 ] + c2*coeff[index2 ] + bias) >> COL_SHIFT;
400 block[i+8 ]= (c1*coeff[index1+1] + c2*coeff[index2+1] + bias) >> COL_SHIFT;
401 block[i+16]= (c1*coeff[index1+2] + c2*coeff[index2+2] + bias) >> COL_SHIFT;
402 block[i+24]= (c1*coeff[index1+3] + c2*coeff[index2+3] + bias) >> COL_SHIFT;
403 block[i+32]= (c1*coeff[index1+4] + c2*coeff[index2+4] + bias) >> COL_SHIFT;
404 block[i+40]= (c1*coeff[index1+5] + c2*coeff[index2+5] + bias) >> COL_SHIFT;
405 block[i+48]= (c1*coeff[index1+6] + c2*coeff[index2+6] + bias) >> COL_SHIFT;
406 block[i+56]= (c1*coeff[index1+7] + c2*coeff[index2+7] + bias) >> COL_SHIFT;
407 }
408 }*/
409 else
410 {
411 for(i=0; i<8; i++)
412 idctSparse2Col(block + i);
413 }
414 #else
223 for(i=0; i<8; i++) 415 for(i=0; i<8; i++)
224 idctRow(temp+8*i, block + 8*i); 416 idctRowCondDC(block + i*8);
225 417
226 for(i=0; i<8; i++) 418 for(i=0; i<8; i++)
227 idctCol(temp+i, block + i); 419 idctSparseCol(block + i);
228
229 }
230
231 #endif 420 #endif
421 }