libavcodec/vc1dsp.c
Go to the documentation of this file.
00001 /*
00002  * VC-1 and WMV3 decoder - DSP functions
00003  * Copyright (c) 2006 Konstantin Shishkov
00004  *
00005  * This file is part of Libav.
00006  *
00007  * Libav is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * Libav is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with Libav; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00028 #include "vc1dsp.h"
00029 
00030 
00033 static void vc1_v_overlap_c(uint8_t* src, int stride)
00034 {
00035     int i;
00036     int a, b, c, d;
00037     int d1, d2;
00038     int rnd = 1;
00039     for(i = 0; i < 8; i++) {
00040         a = src[-2*stride];
00041         b = src[-stride];
00042         c = src[0];
00043         d = src[stride];
00044         d1 = (a - d + 3 + rnd) >> 3;
00045         d2 = (a - d + b - c + 4 - rnd) >> 3;
00046 
00047         src[-2*stride] = a - d1;
00048         src[-stride] = av_clip_uint8(b - d2);
00049         src[0] = av_clip_uint8(c + d2);
00050         src[stride] = d + d1;
00051         src++;
00052         rnd = !rnd;
00053     }
00054 }
00055 
00058 static void vc1_h_overlap_c(uint8_t* src, int stride)
00059 {
00060     int i;
00061     int a, b, c, d;
00062     int d1, d2;
00063     int rnd = 1;
00064     for(i = 0; i < 8; i++) {
00065         a = src[-2];
00066         b = src[-1];
00067         c = src[0];
00068         d = src[1];
00069         d1 = (a - d + 3 + rnd) >> 3;
00070         d2 = (a - d + b - c + 4 - rnd) >> 3;
00071 
00072         src[-2] = a - d1;
00073         src[-1] = av_clip_uint8(b - d2);
00074         src[0] = av_clip_uint8(c + d2);
00075         src[1] = d + d1;
00076         src += stride;
00077         rnd = !rnd;
00078     }
00079 }
00080 
00081 static void vc1_v_s_overlap_c(DCTELEM *top,  DCTELEM *bottom)
00082 {
00083     int i;
00084     int a, b, c, d;
00085     int d1, d2;
00086     int rnd1 = 4, rnd2 = 3;
00087     for(i = 0; i < 8; i++) {
00088         a = top[48];
00089         b = top[56];
00090         c = bottom[0];
00091         d = bottom[8];
00092         d1 = a - d;
00093         d2 = a - d + b - c;
00094 
00095         top[48]   = ((a << 3) - d1 + rnd1) >> 3;
00096         top[56]   = ((b << 3) - d2 + rnd2) >> 3;
00097         bottom[0] = ((c << 3) + d2 + rnd1) >> 3;
00098         bottom[8] = ((d << 3) + d1 + rnd2) >> 3;
00099 
00100         bottom++;
00101         top++;
00102         rnd2 = 7 - rnd2;
00103         rnd1 = 7 - rnd1;
00104     }
00105 }
00106 
00107 static void vc1_h_s_overlap_c(DCTELEM *left, DCTELEM *right)
00108 {
00109     int i;
00110     int a, b, c, d;
00111     int d1, d2;
00112     int rnd1 = 4, rnd2 = 3;
00113     for(i = 0; i < 8; i++) {
00114         a = left[6];
00115         b = left[7];
00116         c = right[0];
00117         d = right[1];
00118         d1 = a - d;
00119         d2 = a - d + b - c;
00120 
00121         left[6]  = ((a << 3) - d1 + rnd1) >> 3;
00122         left[7]  = ((b << 3) - d2 + rnd2) >> 3;
00123         right[0] = ((c << 3) + d2 + rnd1) >> 3;
00124         right[1] = ((d << 3) + d1 + rnd2) >> 3;
00125 
00126         right += 8;
00127         left += 8;
00128         rnd2 = 7 - rnd2;
00129         rnd1 = 7 - rnd1;
00130     }
00131 }
00132 
00141 static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
00142     int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
00143     int a0_sign = a0 >> 31;        /* Store sign */
00144     a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
00145     if(a0 < pq){
00146         int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
00147         int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
00148         if(a1 < a0 || a2 < a0){
00149             int clip = src[-1*stride] - src[ 0*stride];
00150             int clip_sign = clip >> 31;
00151             clip = ((clip ^ clip_sign) - clip_sign)>>1;
00152             if(clip){
00153                 int a3 = FFMIN(a1, a2);
00154                 int d = 5 * (a3 - a0);
00155                 int d_sign = (d >> 31);
00156                 d = ((d ^ d_sign) - d_sign) >> 3;
00157                 d_sign ^= a0_sign;
00158 
00159                 if( d_sign ^ clip_sign )
00160                     d = 0;
00161                 else{
00162                     d = FFMIN(d, clip);
00163                     d = (d ^ d_sign) - d_sign;          /* Restore sign */
00164                     src[-1*stride] = av_clip_uint8(src[-1*stride] - d);
00165                     src[ 0*stride] = av_clip_uint8(src[ 0*stride] + d);
00166                 }
00167                 return 1;
00168             }
00169         }
00170     }
00171     return 0;
00172 }
00173 
00183 static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
00184 {
00185     int i;
00186     int filt3;
00187 
00188     for(i = 0; i < len; i += 4){
00189         filt3 = vc1_filter_line(src + 2*step, stride, pq);
00190         if(filt3){
00191             vc1_filter_line(src + 0*step, stride, pq);
00192             vc1_filter_line(src + 1*step, stride, pq);
00193             vc1_filter_line(src + 3*step, stride, pq);
00194         }
00195         src += step * 4;
00196     }
00197 }
00198 
00199 static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq)
00200 {
00201     vc1_loop_filter(src, 1, stride, 4, pq);
00202 }
00203 
00204 static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq)
00205 {
00206     vc1_loop_filter(src, stride, 1, 4, pq);
00207 }
00208 
00209 static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq)
00210 {
00211     vc1_loop_filter(src, 1, stride, 8, pq);
00212 }
00213 
00214 static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq)
00215 {
00216     vc1_loop_filter(src, stride, 1, 8, pq);
00217 }
00218 
00219 static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq)
00220 {
00221     vc1_loop_filter(src, 1, stride, 16, pq);
00222 }
00223 
00224 static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)
00225 {
00226     vc1_loop_filter(src, stride, 1, 16, pq);
00227 }
00228 
00231 static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
00232 {
00233     int i;
00234     int dc = block[0];
00235     dc = (3 * dc +  1) >> 1;
00236     dc = (3 * dc + 16) >> 5;
00237     for(i = 0; i < 8; i++){
00238         dest[0] = av_clip_uint8(dest[0] + dc);
00239         dest[1] = av_clip_uint8(dest[1] + dc);
00240         dest[2] = av_clip_uint8(dest[2] + dc);
00241         dest[3] = av_clip_uint8(dest[3] + dc);
00242         dest[4] = av_clip_uint8(dest[4] + dc);
00243         dest[5] = av_clip_uint8(dest[5] + dc);
00244         dest[6] = av_clip_uint8(dest[6] + dc);
00245         dest[7] = av_clip_uint8(dest[7] + dc);
00246         dest += linesize;
00247     }
00248 }
00249 
00250 static void vc1_inv_trans_8x8_c(DCTELEM block[64])
00251 {
00252     int i;
00253     register int t1,t2,t3,t4,t5,t6,t7,t8;
00254     DCTELEM *src, *dst, temp[64];
00255 
00256     src = block;
00257     dst = temp;
00258     for(i = 0; i < 8; i++){
00259         t1 = 12 * (src[ 0] + src[32]) + 4;
00260         t2 = 12 * (src[ 0] - src[32]) + 4;
00261         t3 = 16 * src[16] +  6 * src[48];
00262         t4 =  6 * src[16] - 16 * src[48];
00263 
00264         t5 = t1 + t3;
00265         t6 = t2 + t4;
00266         t7 = t2 - t4;
00267         t8 = t1 - t3;
00268 
00269         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
00270         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
00271         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
00272         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
00273 
00274         dst[0] = (t5 + t1) >> 3;
00275         dst[1] = (t6 + t2) >> 3;
00276         dst[2] = (t7 + t3) >> 3;
00277         dst[3] = (t8 + t4) >> 3;
00278         dst[4] = (t8 - t4) >> 3;
00279         dst[5] = (t7 - t3) >> 3;
00280         dst[6] = (t6 - t2) >> 3;
00281         dst[7] = (t5 - t1) >> 3;
00282 
00283         src += 1;
00284         dst += 8;
00285     }
00286 
00287     src = temp;
00288     dst = block;
00289     for(i = 0; i < 8; i++){
00290         t1 = 12 * (src[ 0] + src[32]) + 64;
00291         t2 = 12 * (src[ 0] - src[32]) + 64;
00292         t3 = 16 * src[16] +  6 * src[48];
00293         t4 =  6 * src[16] - 16 * src[48];
00294 
00295         t5 = t1 + t3;
00296         t6 = t2 + t4;
00297         t7 = t2 - t4;
00298         t8 = t1 - t3;
00299 
00300         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
00301         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
00302         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
00303         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
00304 
00305         dst[ 0] = (t5 + t1) >> 7;
00306         dst[ 8] = (t6 + t2) >> 7;
00307         dst[16] = (t7 + t3) >> 7;
00308         dst[24] = (t8 + t4) >> 7;
00309         dst[32] = (t8 - t4 + 1) >> 7;
00310         dst[40] = (t7 - t3 + 1) >> 7;
00311         dst[48] = (t6 - t2 + 1) >> 7;
00312         dst[56] = (t5 - t1 + 1) >> 7;
00313 
00314         src++;
00315         dst++;
00316     }
00317 }
00318 
00321 static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
00322 {
00323     int i;
00324     int dc = block[0];
00325     dc = ( 3 * dc +  1) >> 1;
00326     dc = (17 * dc + 64) >> 7;
00327     for(i = 0; i < 4; i++){
00328         dest[0] = av_clip_uint8(dest[0] + dc);
00329         dest[1] = av_clip_uint8(dest[1] + dc);
00330         dest[2] = av_clip_uint8(dest[2] + dc);
00331         dest[3] = av_clip_uint8(dest[3] + dc);
00332         dest[4] = av_clip_uint8(dest[4] + dc);
00333         dest[5] = av_clip_uint8(dest[5] + dc);
00334         dest[6] = av_clip_uint8(dest[6] + dc);
00335         dest[7] = av_clip_uint8(dest[7] + dc);
00336         dest += linesize;
00337     }
00338 }
00339 
00340 static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
00341 {
00342     int i;
00343     register int t1,t2,t3,t4,t5,t6,t7,t8;
00344     DCTELEM *src, *dst;
00345 
00346     src = block;
00347     dst = block;
00348     for(i = 0; i < 4; i++){
00349         t1 = 12 * (src[0] + src[4]) + 4;
00350         t2 = 12 * (src[0] - src[4]) + 4;
00351         t3 = 16 * src[2] +  6 * src[6];
00352         t4 =  6 * src[2] - 16 * src[6];
00353 
00354         t5 = t1 + t3;
00355         t6 = t2 + t4;
00356         t7 = t2 - t4;
00357         t8 = t1 - t3;
00358 
00359         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
00360         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
00361         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
00362         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
00363 
00364         dst[0] = (t5 + t1) >> 3;
00365         dst[1] = (t6 + t2) >> 3;
00366         dst[2] = (t7 + t3) >> 3;
00367         dst[3] = (t8 + t4) >> 3;
00368         dst[4] = (t8 - t4) >> 3;
00369         dst[5] = (t7 - t3) >> 3;
00370         dst[6] = (t6 - t2) >> 3;
00371         dst[7] = (t5 - t1) >> 3;
00372 
00373         src += 8;
00374         dst += 8;
00375     }
00376 
00377     src = block;
00378     for(i = 0; i < 8; i++){
00379         t1 = 17 * (src[ 0] + src[16]) + 64;
00380         t2 = 17 * (src[ 0] - src[16]) + 64;
00381         t3 = 22 * src[ 8] + 10 * src[24];
00382         t4 = 22 * src[24] - 10 * src[ 8];
00383 
00384         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));
00385         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));
00386         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));
00387         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));
00388 
00389         src ++;
00390         dest++;
00391     }
00392 }
00393 
00396 static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
00397 {
00398     int i;
00399     int dc = block[0];
00400     dc = (17 * dc +  4) >> 3;
00401     dc = (12 * dc + 64) >> 7;
00402     for(i = 0; i < 8; i++){
00403         dest[0] = av_clip_uint8(dest[0] + dc);
00404         dest[1] = av_clip_uint8(dest[1] + dc);
00405         dest[2] = av_clip_uint8(dest[2] + dc);
00406         dest[3] = av_clip_uint8(dest[3] + dc);
00407         dest += linesize;
00408     }
00409 }
00410 
00411 static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
00412 {
00413     int i;
00414     register int t1,t2,t3,t4,t5,t6,t7,t8;
00415     DCTELEM *src, *dst;
00416 
00417     src = block;
00418     dst = block;
00419     for(i = 0; i < 8; i++){
00420         t1 = 17 * (src[0] + src[2]) + 4;
00421         t2 = 17 * (src[0] - src[2]) + 4;
00422         t3 = 22 * src[1] + 10 * src[3];
00423         t4 = 22 * src[3] - 10 * src[1];
00424 
00425         dst[0] = (t1 + t3) >> 3;
00426         dst[1] = (t2 - t4) >> 3;
00427         dst[2] = (t2 + t4) >> 3;
00428         dst[3] = (t1 - t3) >> 3;
00429 
00430         src += 8;
00431         dst += 8;
00432     }
00433 
00434     src = block;
00435     for(i = 0; i < 4; i++){
00436         t1 = 12 * (src[ 0] + src[32]) + 64;
00437         t2 = 12 * (src[ 0] - src[32]) + 64;
00438         t3 = 16 * src[16] +  6 * src[48];
00439         t4 =  6 * src[16] - 16 * src[48];
00440 
00441         t5 = t1 + t3;
00442         t6 = t2 + t4;
00443         t7 = t2 - t4;
00444         t8 = t1 - t3;
00445 
00446         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
00447         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
00448         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
00449         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
00450 
00451         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t5 + t1) >> 7));
00452         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t6 + t2) >> 7));
00453         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t7 + t3) >> 7));
00454         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t8 + t4) >> 7));
00455         dest[4*linesize] = av_clip_uint8(dest[4*linesize] + ((t8 - t4 + 1) >> 7));
00456         dest[5*linesize] = av_clip_uint8(dest[5*linesize] + ((t7 - t3 + 1) >> 7));
00457         dest[6*linesize] = av_clip_uint8(dest[6*linesize] + ((t6 - t2 + 1) >> 7));
00458         dest[7*linesize] = av_clip_uint8(dest[7*linesize] + ((t5 - t1 + 1) >> 7));
00459 
00460         src ++;
00461         dest++;
00462     }
00463 }
00464 
00467 static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
00468 {
00469     int i;
00470     int dc = block[0];
00471     dc = (17 * dc +  4) >> 3;
00472     dc = (17 * dc + 64) >> 7;
00473     for(i = 0; i < 4; i++){
00474         dest[0] = av_clip_uint8(dest[0] + dc);
00475         dest[1] = av_clip_uint8(dest[1] + dc);
00476         dest[2] = av_clip_uint8(dest[2] + dc);
00477         dest[3] = av_clip_uint8(dest[3] + dc);
00478         dest += linesize;
00479     }
00480 }
00481 
00482 static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
00483 {
00484     int i;
00485     register int t1,t2,t3,t4;
00486     DCTELEM *src, *dst;
00487 
00488     src = block;
00489     dst = block;
00490     for(i = 0; i < 4; i++){
00491         t1 = 17 * (src[0] + src[2]) + 4;
00492         t2 = 17 * (src[0] - src[2]) + 4;
00493         t3 = 22 * src[1] + 10 * src[3];
00494         t4 = 22 * src[3] - 10 * src[1];
00495 
00496         dst[0] = (t1 + t3) >> 3;
00497         dst[1] = (t2 - t4) >> 3;
00498         dst[2] = (t2 + t4) >> 3;
00499         dst[3] = (t1 - t3) >> 3;
00500 
00501         src += 8;
00502         dst += 8;
00503     }
00504 
00505     src = block;
00506     for(i = 0; i < 4; i++){
00507         t1 = 17 * (src[ 0] + src[16]) + 64;
00508         t2 = 17 * (src[ 0] - src[16]) + 64;
00509         t3 = 22 * src[ 8] + 10 * src[24];
00510         t4 = 22 * src[24] - 10 * src[ 8];
00511 
00512         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));
00513         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));
00514         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));
00515         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));
00516 
00517         src ++;
00518         dest++;
00519     }
00520 }
00521 
00522 /* motion compensation functions */
00524 #define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \
00525 static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \
00526 {                                                                       \
00527     switch(mode){                                                       \
00528     case 0: /* no shift - should not occur */                           \
00529         return 0;                                                       \
00530     case 1: /* 1/4 shift */                                             \
00531         return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \
00532     case 2: /* 1/2 shift */                                             \
00533         return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \
00534     case 3: /* 3/4 shift */                                             \
00535         return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \
00536     }                                                                   \
00537     return 0; /* should not occur */                                    \
00538 }
00539 
00540 VC1_MSPEL_FILTER_16B(ver, uint8_t)
00541 VC1_MSPEL_FILTER_16B(hor, int16_t)
00542 
00543 
00546 static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
00547 {
00548     switch(mode){
00549     case 0: //no shift
00550         return src[0];
00551     case 1: // 1/4 shift
00552         return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
00553     case 2: // 1/2 shift
00554         return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
00555     case 3: // 3/4 shift
00556         return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
00557     }
00558     return 0; //should not occur
00559 }
00560 
00563 #define VC1_MSPEL_MC(OP, OPNAME)\
00564 static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
00565 {\
00566     int     i, j;\
00567 \
00568     if (vmode) { /* Horizontal filter to apply */\
00569         int r;\
00570 \
00571         if (hmode) { /* Vertical filter to apply, output to tmp */\
00572             static const int shift_value[] = { 0, 5, 1, 5 };\
00573             int              shift = (shift_value[hmode]+shift_value[vmode])>>1;\
00574             int16_t          tmp[11*8], *tptr = tmp;\
00575 \
00576             r = (1<<(shift-1)) + rnd-1;\
00577 \
00578             src -= 1;\
00579             for(j = 0; j < 8; j++) {\
00580                 for(i = 0; i < 11; i++)\
00581                     tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\
00582                 src += stride;\
00583                 tptr += 11;\
00584             }\
00585 \
00586             r = 64-rnd;\
00587             tptr = tmp+1;\
00588             for(j = 0; j < 8; j++) {\
00589                 for(i = 0; i < 8; i++)\
00590                     OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\
00591                 dst += stride;\
00592                 tptr += 11;\
00593             }\
00594 \
00595             return;\
00596         }\
00597         else { /* No horizontal filter, output 8 lines to dst */\
00598             r = 1-rnd;\
00599 \
00600             for(j = 0; j < 8; j++) {\
00601                 for(i = 0; i < 8; i++)\
00602                     OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\
00603                 src += stride;\
00604                 dst += stride;\
00605             }\
00606             return;\
00607         }\
00608     }\
00609 \
00610     /* Horizontal mode with no vertical mode */\
00611     for(j = 0; j < 8; j++) {\
00612         for(i = 0; i < 8; i++)\
00613             OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\
00614         dst += stride;\
00615         src += stride;\
00616     }\
00617 }
00618 
00619 #define op_put(a, b) a = av_clip_uint8(b)
00620 #define op_avg(a, b) a = (a + av_clip_uint8(b) + 1) >> 1
00621 
00622 VC1_MSPEL_MC(op_put, put_)
00623 VC1_MSPEL_MC(op_avg, avg_)
00624 
00625 /* pixel functions - really are entry points to vc1_mspel_mc */
00626 
00627 #define PUT_VC1_MSPEL(a, b)\
00628 static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
00629      put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
00630 }\
00631 static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
00632      avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
00633 }
00634 
00635 PUT_VC1_MSPEL(1, 0)
00636 PUT_VC1_MSPEL(2, 0)
00637 PUT_VC1_MSPEL(3, 0)
00638 
00639 PUT_VC1_MSPEL(0, 1)
00640 PUT_VC1_MSPEL(1, 1)
00641 PUT_VC1_MSPEL(2, 1)
00642 PUT_VC1_MSPEL(3, 1)
00643 
00644 PUT_VC1_MSPEL(0, 2)
00645 PUT_VC1_MSPEL(1, 2)
00646 PUT_VC1_MSPEL(2, 2)
00647 PUT_VC1_MSPEL(3, 2)
00648 
00649 PUT_VC1_MSPEL(0, 3)
00650 PUT_VC1_MSPEL(1, 3)
00651 PUT_VC1_MSPEL(2, 3)
00652 PUT_VC1_MSPEL(3, 3)
00653 
00654 static void put_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){
00655     const int A=(8-x)*(8-y);
00656     const int B=(  x)*(8-y);
00657     const int C=(8-x)*(  y);
00658     const int D=(  x)*(  y);
00659     int i;
00660 
00661     assert(x<8 && y<8 && x>=0 && y>=0);
00662 
00663     for(i=0; i<h; i++)
00664     {
00665         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;
00666         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;
00667         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;
00668         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;
00669         dst[4] = (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6;
00670         dst[5] = (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6;
00671         dst[6] = (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6;
00672         dst[7] = (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6;
00673         dst+= stride;
00674         src+= stride;
00675     }
00676 }
00677 
00678 static void put_no_rnd_vc1_chroma_mc4_c(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y){
00679     const int A=(8-x)*(8-y);
00680     const int B=(  x)*(8-y);
00681     const int C=(8-x)*(  y);
00682     const int D=(  x)*(  y);
00683     int i;
00684 
00685     assert(x<8 && y<8 && x>=0 && y>=0);
00686 
00687     for(i=0; i<h; i++)
00688     {
00689         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;
00690         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;
00691         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;
00692         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;
00693         dst+= stride;
00694         src+= stride;
00695     }
00696 }
00697 
00698 #define avg2(a,b) ((a+b+1)>>1)
00699 static void avg_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){
00700     const int A=(8-x)*(8-y);
00701     const int B=(  x)*(8-y);
00702     const int C=(8-x)*(  y);
00703     const int D=(  x)*(  y);
00704     int i;
00705 
00706     assert(x<8 && y<8 && x>=0 && y>=0);
00707 
00708     for(i=0; i<h; i++)
00709     {
00710         dst[0] = avg2(dst[0], ((A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6));
00711         dst[1] = avg2(dst[1], ((A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6));
00712         dst[2] = avg2(dst[2], ((A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6));
00713         dst[3] = avg2(dst[3], ((A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6));
00714         dst[4] = avg2(dst[4], ((A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6));
00715         dst[5] = avg2(dst[5], ((A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6));
00716         dst[6] = avg2(dst[6], ((A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6));
00717         dst[7] = avg2(dst[7], ((A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6));
00718         dst+= stride;
00719         src+= stride;
00720     }
00721 }
00722 
00723 #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
00724 
00725 static void sprite_h_c(uint8_t *dst, const uint8_t *src, int offset, int advance, int count)
00726 {
00727     while (count--) {
00728         int a = src[(offset >> 16)    ];
00729         int b = src[(offset >> 16) + 1];
00730         *dst++ = a + ((b - a) * (offset&0xFFFF) >> 16);
00731         offset += advance;
00732     }
00733 }
00734 
00735 static av_always_inline void sprite_v_template(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
00736                                             int two_sprites, const uint8_t *src2a, const uint8_t *src2b, int offset2,
00737                                             int alpha, int scaled, int width)
00738 {
00739     int a1, b1, a2, b2;
00740     while (width--) {
00741         a1 = *src1a++;
00742         if (scaled) {
00743             b1 = *src1b++;
00744             a1 = a1 + ((b1 - a1) * offset1 >> 16);
00745         }
00746         if (two_sprites) {
00747             a2 = *src2a++;
00748             if (scaled > 1) {
00749                 b2 = *src2b++;
00750                 a2 = a2 + ((b2 - a2) * offset2 >> 16);
00751             }
00752             a1 = a1 + ((a2 - a1) * alpha >> 16);
00753         }
00754         *dst++ = a1;
00755     }
00756 }
00757 
00758 static void sprite_v_single_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset, int width)
00759 {
00760     sprite_v_template(dst, src1a, src1b, offset, 0, NULL, NULL, 0, 0, 1, width);
00761 }
00762 
00763 static void sprite_v_double_noscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src2a, int alpha, int width)
00764 {
00765     sprite_v_template(dst, src1a, NULL, 0, 1, src2a, NULL, 0, alpha, 0, width);
00766 }
00767 
00768 static void sprite_v_double_onescale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
00769                                                      const uint8_t *src2a, int alpha, int width)
00770 {
00771     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, NULL, 0, alpha, 1, width);
00772 }
00773 
00774 static void sprite_v_double_twoscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
00775                                                      const uint8_t *src2a, const uint8_t *src2b, int offset2,
00776                                        int alpha, int width)
00777 {
00778     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, src2b, offset2, alpha, 2, width);
00779 }
00780 
00781 #endif
00782 
00783 av_cold void ff_vc1dsp_init(VC1DSPContext* dsp) {
00784     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
00785     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
00786     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
00787     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
00788     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;
00789     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;
00790     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;
00791     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;
00792     dsp->vc1_h_overlap = vc1_h_overlap_c;
00793     dsp->vc1_v_overlap = vc1_v_overlap_c;
00794     dsp->vc1_h_s_overlap = vc1_h_s_overlap_c;
00795     dsp->vc1_v_s_overlap = vc1_v_s_overlap_c;
00796     dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
00797     dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
00798     dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;
00799     dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;
00800     dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;
00801     dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;
00802 
00803     dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_pixels8x8_c;
00804     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
00805     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
00806     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
00807     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
00808     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
00809     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
00810     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
00811     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
00812     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
00813     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
00814     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
00815     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
00816     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
00817     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
00818     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
00819 
00820     dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_pixels8x8_c;
00821     dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c;
00822     dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c;
00823     dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c;
00824     dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c;
00825     dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c;
00826     dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c;
00827     dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c;
00828     dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c;
00829     dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c;
00830     dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;
00831     dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;
00832     dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;
00833     dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;
00834     dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;
00835     dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;
00836 
00837     dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= put_no_rnd_vc1_chroma_mc8_c;
00838     dsp->avg_no_rnd_vc1_chroma_pixels_tab[0]= avg_no_rnd_vc1_chroma_mc8_c;
00839     dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = put_no_rnd_vc1_chroma_mc4_c;
00840 
00841 #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
00842     dsp->sprite_h = sprite_h_c;
00843     dsp->sprite_v_single = sprite_v_single_c;
00844     dsp->sprite_v_double_noscale = sprite_v_double_noscale_c;
00845     dsp->sprite_v_double_onescale = sprite_v_double_onescale_c;
00846     dsp->sprite_v_double_twoscale = sprite_v_double_twoscale_c;
00847 #endif
00848 
00849     if (HAVE_ALTIVEC)
00850         ff_vc1dsp_init_altivec(dsp);
00851     if (HAVE_MMX)
00852         ff_vc1dsp_init_mmx(dsp);
00853 }