yading@10: /*
yading@10:  * VC-1 and WMV3 decoder - DSP functions
yading@10:  * Copyright (c) 2006 Konstantin Shishkov
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10: * @file
yading@10:  * VC-1 and WMV3 decoder
yading@10:  *
yading@10:  */
yading@10: 
yading@10: #include "libavutil/avassert.h"
yading@10: #include "libavutil/common.h"
yading@10: #include "h264chroma.h"
yading@10: #include "rnd_avg.h"
yading@10: #include "vc1dsp.h"
yading@10: 
yading@10: 
yading@10: /** Apply overlap transform to horizontal edge
yading@10: */
yading@10: static void vc1_v_overlap_c(uint8_t* src, int stride)
yading@10: {
yading@10:     int i;
yading@10:     int a, b, c, d;
yading@10:     int d1, d2;
yading@10:     int rnd = 1;
yading@10:     for(i = 0; i < 8; i++) {
yading@10:         a = src[-2*stride];
yading@10:         b = src[-stride];
yading@10:         c = src[0];
yading@10:         d = src[stride];
yading@10:         d1 = (a - d + 3 + rnd) >> 3;
yading@10:         d2 = (a - d + b - c + 4 - rnd) >> 3;
yading@10: 
yading@10:         src[-2*stride] = a - d1;
yading@10:         src[-stride] = av_clip_uint8(b - d2);
yading@10:         src[0] = av_clip_uint8(c + d2);
yading@10:         src[stride] = d + d1;
yading@10:         src++;
yading@10:         rnd = !rnd;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /** Apply overlap transform to vertical edge
yading@10: */
yading@10: static void vc1_h_overlap_c(uint8_t* src, int stride)
yading@10: {
yading@10:     int i;
yading@10:     int a, b, c, d;
yading@10:     int d1, d2;
yading@10:     int rnd = 1;
yading@10:     for(i = 0; i < 8; i++) {
yading@10:         a = src[-2];
yading@10:         b = src[-1];
yading@10:         c = src[0];
yading@10:         d = src[1];
yading@10:         d1 = (a - d + 3 + rnd) >> 3;
yading@10:         d2 = (a - d + b - c + 4 - rnd) >> 3;
yading@10: 
yading@10:         src[-2] = a - d1;
yading@10:         src[-1] = av_clip_uint8(b - d2);
yading@10:         src[0] = av_clip_uint8(c + d2);
yading@10:         src[1] = d + d1;
yading@10:         src += stride;
yading@10:         rnd = !rnd;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_v_s_overlap_c(int16_t *top,  int16_t *bottom)
yading@10: {
yading@10:     int i;
yading@10:     int a, b, c, d;
yading@10:     int d1, d2;
yading@10:     int rnd1 = 4, rnd2 = 3;
yading@10:     for(i = 0; i < 8; i++) {
yading@10:         a = top[48];
yading@10:         b = top[56];
yading@10:         c = bottom[0];
yading@10:         d = bottom[8];
yading@10:         d1 = a - d;
yading@10:         d2 = a - d + b - c;
yading@10: 
yading@10:         top[48]   = ((a << 3) - d1 + rnd1) >> 3;
yading@10:         top[56]   = ((b << 3) - d2 + rnd2) >> 3;
yading@10:         bottom[0] = ((c << 3) + d2 + rnd1) >> 3;
yading@10:         bottom[8] = ((d << 3) + d1 + rnd2) >> 3;
yading@10: 
yading@10:         bottom++;
yading@10:         top++;
yading@10:         rnd2 = 7 - rnd2;
yading@10:         rnd1 = 7 - rnd1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_h_s_overlap_c(int16_t *left, int16_t *right)
yading@10: {
yading@10:     int i;
yading@10:     int a, b, c, d;
yading@10:     int d1, d2;
yading@10:     int rnd1 = 4, rnd2 = 3;
yading@10:     for(i = 0; i < 8; i++) {
yading@10:         a = left[6];
yading@10:         b = left[7];
yading@10:         c = right[0];
yading@10:         d = right[1];
yading@10:         d1 = a - d;
yading@10:         d2 = a - d + b - c;
yading@10: 
yading@10:         left[6]  = ((a << 3) - d1 + rnd1) >> 3;
yading@10:         left[7]  = ((b << 3) - d2 + rnd2) >> 3;
yading@10:         right[0] = ((c << 3) + d2 + rnd1) >> 3;
yading@10:         right[1] = ((d << 3) + d1 + rnd2) >> 3;
yading@10: 
yading@10:         right += 8;
yading@10:         left += 8;
yading@10:         rnd2 = 7 - rnd2;
yading@10:         rnd1 = 7 - rnd1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * VC-1 in-loop deblocking filter for one line
yading@10:  * @param src source block type
yading@10:  * @param stride block stride
yading@10:  * @param pq block quantizer
yading@10:  * @return whether other 3 pairs should be filtered or not
yading@10:  * @see 8.6
yading@10:  */
yading@10: static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
yading@10:     int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
yading@10:     int a0_sign = a0 >> 31;        /* Store sign */
yading@10:     a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
yading@10:     if(a0 < pq){
yading@10:         int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
yading@10:         int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
yading@10:         if(a1 < a0 || a2 < a0){
yading@10:             int clip = src[-1*stride] - src[ 0*stride];
yading@10:             int clip_sign = clip >> 31;
yading@10:             clip = ((clip ^ clip_sign) - clip_sign)>>1;
yading@10:             if(clip){
yading@10:                 int a3 = FFMIN(a1, a2);
yading@10:                 int d = 5 * (a3 - a0);
yading@10:                 int d_sign = (d >> 31);
yading@10:                 d = ((d ^ d_sign) - d_sign) >> 3;
yading@10:                 d_sign ^= a0_sign;
yading@10: 
yading@10:                 if( d_sign ^ clip_sign )
yading@10:                     d = 0;
yading@10:                 else{
yading@10:                     d = FFMIN(d, clip);
yading@10:                     d = (d ^ d_sign) - d_sign;          /* Restore sign */
yading@10:                     src[-1*stride] = av_clip_uint8(src[-1*stride] - d);
yading@10:                     src[ 0*stride] = av_clip_uint8(src[ 0*stride] + d);
yading@10:                 }
yading@10:                 return 1;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * VC-1 in-loop deblocking filter
yading@10:  * @param src source block type
yading@10:  * @param step distance between horizontally adjacent elements
yading@10:  * @param stride distance between vertically adjacent elements
yading@10:  * @param len edge length to filter (4 or 8 pixels)
yading@10:  * @param pq block quantizer
yading@10:  * @see 8.6
yading@10:  */
yading@10: static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
yading@10: {
yading@10:     int i;
yading@10:     int filt3;
yading@10: 
yading@10:     for(i = 0; i < len; i += 4){
yading@10:         filt3 = vc1_filter_line(src + 2*step, stride, pq);
yading@10:         if(filt3){
yading@10:             vc1_filter_line(src + 0*step, stride, pq);
yading@10:             vc1_filter_line(src + 1*step, stride, pq);
yading@10:             vc1_filter_line(src + 3*step, stride, pq);
yading@10:         }
yading@10:         src += step * 4;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, 1, stride, 4, pq);
yading@10: }
yading@10: 
yading@10: static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, stride, 1, 4, pq);
yading@10: }
yading@10: 
yading@10: static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, 1, stride, 8, pq);
yading@10: }
yading@10: 
yading@10: static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, stride, 1, 8, pq);
yading@10: }
yading@10: 
yading@10: static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, 1, stride, 16, pq);
yading@10: }
yading@10: 
yading@10: static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)
yading@10: {
yading@10:     vc1_loop_filter(src, stride, 1, 16, pq);
yading@10: }
yading@10: 
yading@10: /** Do inverse transform on 8x8 block
yading@10: */
yading@10: static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     int dc = block[0];
yading@10:     dc = (3 * dc +  1) >> 1;
yading@10:     dc = (3 * dc + 16) >> 5;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         dest[0] = av_clip_uint8(dest[0] + dc);
yading@10:         dest[1] = av_clip_uint8(dest[1] + dc);
yading@10:         dest[2] = av_clip_uint8(dest[2] + dc);
yading@10:         dest[3] = av_clip_uint8(dest[3] + dc);
yading@10:         dest[4] = av_clip_uint8(dest[4] + dc);
yading@10:         dest[5] = av_clip_uint8(dest[5] + dc);
yading@10:         dest[6] = av_clip_uint8(dest[6] + dc);
yading@10:         dest[7] = av_clip_uint8(dest[7] + dc);
yading@10:         dest += linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_inv_trans_8x8_c(int16_t block[64])
yading@10: {
yading@10:     int i;
yading@10:     register int t1,t2,t3,t4,t5,t6,t7,t8;
yading@10:     int16_t *src, *dst, temp[64];
yading@10: 
yading@10:     src = block;
yading@10:     dst = temp;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         t1 = 12 * (src[ 0] + src[32]) + 4;
yading@10:         t2 = 12 * (src[ 0] - src[32]) + 4;
yading@10:         t3 = 16 * src[16] +  6 * src[48];
yading@10:         t4 =  6 * src[16] - 16 * src[48];
yading@10: 
yading@10:         t5 = t1 + t3;
yading@10:         t6 = t2 + t4;
yading@10:         t7 = t2 - t4;
yading@10:         t8 = t1 - t3;
yading@10: 
yading@10:         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
yading@10:         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
yading@10:         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
yading@10:         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
yading@10: 
yading@10:         dst[0] = (t5 + t1) >> 3;
yading@10:         dst[1] = (t6 + t2) >> 3;
yading@10:         dst[2] = (t7 + t3) >> 3;
yading@10:         dst[3] = (t8 + t4) >> 3;
yading@10:         dst[4] = (t8 - t4) >> 3;
yading@10:         dst[5] = (t7 - t3) >> 3;
yading@10:         dst[6] = (t6 - t2) >> 3;
yading@10:         dst[7] = (t5 - t1) >> 3;
yading@10: 
yading@10:         src += 1;
yading@10:         dst += 8;
yading@10:     }
yading@10: 
yading@10:     src = temp;
yading@10:     dst = block;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         t1 = 12 * (src[ 0] + src[32]) + 64;
yading@10:         t2 = 12 * (src[ 0] - src[32]) + 64;
yading@10:         t3 = 16 * src[16] +  6 * src[48];
yading@10:         t4 =  6 * src[16] - 16 * src[48];
yading@10: 
yading@10:         t5 = t1 + t3;
yading@10:         t6 = t2 + t4;
yading@10:         t7 = t2 - t4;
yading@10:         t8 = t1 - t3;
yading@10: 
yading@10:         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
yading@10:         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
yading@10:         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
yading@10:         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
yading@10: 
yading@10:         dst[ 0] = (t5 + t1) >> 7;
yading@10:         dst[ 8] = (t6 + t2) >> 7;
yading@10:         dst[16] = (t7 + t3) >> 7;
yading@10:         dst[24] = (t8 + t4) >> 7;
yading@10:         dst[32] = (t8 - t4 + 1) >> 7;
yading@10:         dst[40] = (t7 - t3 + 1) >> 7;
yading@10:         dst[48] = (t6 - t2 + 1) >> 7;
yading@10:         dst[56] = (t5 - t1 + 1) >> 7;
yading@10: 
yading@10:         src++;
yading@10:         dst++;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /** Do inverse transform on 8x4 part of block
yading@10: */
yading@10: static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     int dc = block[0];
yading@10:     dc = ( 3 * dc +  1) >> 1;
yading@10:     dc = (17 * dc + 64) >> 7;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         dest[0] = av_clip_uint8(dest[0] + dc);
yading@10:         dest[1] = av_clip_uint8(dest[1] + dc);
yading@10:         dest[2] = av_clip_uint8(dest[2] + dc);
yading@10:         dest[3] = av_clip_uint8(dest[3] + dc);
yading@10:         dest[4] = av_clip_uint8(dest[4] + dc);
yading@10:         dest[5] = av_clip_uint8(dest[5] + dc);
yading@10:         dest[6] = av_clip_uint8(dest[6] + dc);
yading@10:         dest[7] = av_clip_uint8(dest[7] + dc);
yading@10:         dest += linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     register int t1,t2,t3,t4,t5,t6,t7,t8;
yading@10:     int16_t *src, *dst;
yading@10: 
yading@10:     src = block;
yading@10:     dst = block;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         t1 = 12 * (src[0] + src[4]) + 4;
yading@10:         t2 = 12 * (src[0] - src[4]) + 4;
yading@10:         t3 = 16 * src[2] +  6 * src[6];
yading@10:         t4 =  6 * src[2] - 16 * src[6];
yading@10: 
yading@10:         t5 = t1 + t3;
yading@10:         t6 = t2 + t4;
yading@10:         t7 = t2 - t4;
yading@10:         t8 = t1 - t3;
yading@10: 
yading@10:         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
yading@10:         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
yading@10:         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
yading@10:         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
yading@10: 
yading@10:         dst[0] = (t5 + t1) >> 3;
yading@10:         dst[1] = (t6 + t2) >> 3;
yading@10:         dst[2] = (t7 + t3) >> 3;
yading@10:         dst[3] = (t8 + t4) >> 3;
yading@10:         dst[4] = (t8 - t4) >> 3;
yading@10:         dst[5] = (t7 - t3) >> 3;
yading@10:         dst[6] = (t6 - t2) >> 3;
yading@10:         dst[7] = (t5 - t1) >> 3;
yading@10: 
yading@10:         src += 8;
yading@10:         dst += 8;
yading@10:     }
yading@10: 
yading@10:     src = block;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         t1 = 17 * (src[ 0] + src[16]) + 64;
yading@10:         t2 = 17 * (src[ 0] - src[16]) + 64;
yading@10:         t3 = 22 * src[ 8] + 10 * src[24];
yading@10:         t4 = 22 * src[24] - 10 * src[ 8];
yading@10: 
yading@10:         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));
yading@10:         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));
yading@10:         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));
yading@10:         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));
yading@10: 
yading@10:         src ++;
yading@10:         dest++;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /** Do inverse transform on 4x8 parts of block
yading@10: */
yading@10: static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     int dc = block[0];
yading@10:     dc = (17 * dc +  4) >> 3;
yading@10:     dc = (12 * dc + 64) >> 7;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         dest[0] = av_clip_uint8(dest[0] + dc);
yading@10:         dest[1] = av_clip_uint8(dest[1] + dc);
yading@10:         dest[2] = av_clip_uint8(dest[2] + dc);
yading@10:         dest[3] = av_clip_uint8(dest[3] + dc);
yading@10:         dest += linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     register int t1,t2,t3,t4,t5,t6,t7,t8;
yading@10:     int16_t *src, *dst;
yading@10: 
yading@10:     src = block;
yading@10:     dst = block;
yading@10:     for(i = 0; i < 8; i++){
yading@10:         t1 = 17 * (src[0] + src[2]) + 4;
yading@10:         t2 = 17 * (src[0] - src[2]) + 4;
yading@10:         t3 = 22 * src[1] + 10 * src[3];
yading@10:         t4 = 22 * src[3] - 10 * src[1];
yading@10: 
yading@10:         dst[0] = (t1 + t3) >> 3;
yading@10:         dst[1] = (t2 - t4) >> 3;
yading@10:         dst[2] = (t2 + t4) >> 3;
yading@10:         dst[3] = (t1 - t3) >> 3;
yading@10: 
yading@10:         src += 8;
yading@10:         dst += 8;
yading@10:     }
yading@10: 
yading@10:     src = block;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         t1 = 12 * (src[ 0] + src[32]) + 64;
yading@10:         t2 = 12 * (src[ 0] - src[32]) + 64;
yading@10:         t3 = 16 * src[16] +  6 * src[48];
yading@10:         t4 =  6 * src[16] - 16 * src[48];
yading@10: 
yading@10:         t5 = t1 + t3;
yading@10:         t6 = t2 + t4;
yading@10:         t7 = t2 - t4;
yading@10:         t8 = t1 - t3;
yading@10: 
yading@10:         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
yading@10:         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
yading@10:         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
yading@10:         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
yading@10: 
yading@10:         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t5 + t1) >> 7));
yading@10:         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t6 + t2) >> 7));
yading@10:         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t7 + t3) >> 7));
yading@10:         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t8 + t4) >> 7));
yading@10:         dest[4*linesize] = av_clip_uint8(dest[4*linesize] + ((t8 - t4 + 1) >> 7));
yading@10:         dest[5*linesize] = av_clip_uint8(dest[5*linesize] + ((t7 - t3 + 1) >> 7));
yading@10:         dest[6*linesize] = av_clip_uint8(dest[6*linesize] + ((t6 - t2 + 1) >> 7));
yading@10:         dest[7*linesize] = av_clip_uint8(dest[7*linesize] + ((t5 - t1 + 1) >> 7));
yading@10: 
yading@10:         src ++;
yading@10:         dest++;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /** Do inverse transform on 4x4 part of block
yading@10: */
yading@10: static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     int dc = block[0];
yading@10:     dc = (17 * dc +  4) >> 3;
yading@10:     dc = (17 * dc + 64) >> 7;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         dest[0] = av_clip_uint8(dest[0] + dc);
yading@10:         dest[1] = av_clip_uint8(dest[1] + dc);
yading@10:         dest[2] = av_clip_uint8(dest[2] + dc);
yading@10:         dest[3] = av_clip_uint8(dest[3] + dc);
yading@10:         dest += linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, int16_t *block)
yading@10: {
yading@10:     int i;
yading@10:     register int t1,t2,t3,t4;
yading@10:     int16_t *src, *dst;
yading@10: 
yading@10:     src = block;
yading@10:     dst = block;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         t1 = 17 * (src[0] + src[2]) + 4;
yading@10:         t2 = 17 * (src[0] - src[2]) + 4;
yading@10:         t3 = 22 * src[1] + 10 * src[3];
yading@10:         t4 = 22 * src[3] - 10 * src[1];
yading@10: 
yading@10:         dst[0] = (t1 + t3) >> 3;
yading@10:         dst[1] = (t2 - t4) >> 3;
yading@10:         dst[2] = (t2 + t4) >> 3;
yading@10:         dst[3] = (t1 - t3) >> 3;
yading@10: 
yading@10:         src += 8;
yading@10:         dst += 8;
yading@10:     }
yading@10: 
yading@10:     src = block;
yading@10:     for(i = 0; i < 4; i++){
yading@10:         t1 = 17 * (src[ 0] + src[16]) + 64;
yading@10:         t2 = 17 * (src[ 0] - src[16]) + 64;
yading@10:         t3 = 22 * src[ 8] + 10 * src[24];
yading@10:         t4 = 22 * src[24] - 10 * src[ 8];
yading@10: 
yading@10:         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));
yading@10:         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));
yading@10:         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));
yading@10:         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));
yading@10: 
yading@10:         src ++;
yading@10:         dest++;
yading@10:     }
yading@10: }
yading@10: 
yading@10: /* motion compensation functions */
yading@10: /** Filter in case of 2 filters */
yading@10: #define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \
yading@10: static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \
yading@10: {                                                                       \
yading@10:     switch(mode){                                                       \
yading@10:     case 0: /* no shift - should not occur */                           \
yading@10:         return 0;                                                       \
yading@10:     case 1: /* 1/4 shift */                                             \
yading@10:         return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \
yading@10:     case 2: /* 1/2 shift */                                             \
yading@10:         return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \
yading@10:     case 3: /* 3/4 shift */                                             \
yading@10:         return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \
yading@10:     }                                                                   \
yading@10:     return 0; /* should not occur */                                    \
yading@10: }
yading@10: 
yading@10: VC1_MSPEL_FILTER_16B(ver, uint8_t)
yading@10: VC1_MSPEL_FILTER_16B(hor, int16_t)
yading@10: 
yading@10: 
yading@10: /** Filter used to interpolate fractional pel values
yading@10:  */
yading@10: static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
yading@10: {
yading@10:     switch(mode){
yading@10:     case 0: //no shift
yading@10:         return src[0];
yading@10:     case 1: // 1/4 shift
yading@10:         return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
yading@10:     case 2: // 1/2 shift
yading@10:         return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
yading@10:     case 3: // 3/4 shift
yading@10:         return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
yading@10:     }
yading@10:     return 0; //should not occur
yading@10: }
yading@10: 
yading@10: /** Function used to do motion compensation with bicubic interpolation
yading@10:  */
yading@10: #define VC1_MSPEL_MC(OP, OP4, OPNAME)\
yading@10: static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, ptrdiff_t stride, int hmode, int vmode, int rnd)\
yading@10: {\
yading@10:     int     i, j;\
yading@10: \
yading@10:     if (vmode) { /* Horizontal filter to apply */\
yading@10:         int r;\
yading@10: \
yading@10:         if (hmode) { /* Vertical filter to apply, output to tmp */\
yading@10:             static const int shift_value[] = { 0, 5, 1, 5 };\
yading@10:             int              shift = (shift_value[hmode]+shift_value[vmode])>>1;\
yading@10:             int16_t          tmp[11*8], *tptr = tmp;\
yading@10: \
yading@10:             r = (1<<(shift-1)) + rnd-1;\
yading@10: \
yading@10:             src -= 1;\
yading@10:             for(j = 0; j < 8; j++) {\
yading@10:                 for(i = 0; i < 11; i++)\
yading@10:                     tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\
yading@10:                 src += stride;\
yading@10:                 tptr += 11;\
yading@10:             }\
yading@10: \
yading@10:             r = 64-rnd;\
yading@10:             tptr = tmp+1;\
yading@10:             for(j = 0; j < 8; j++) {\
yading@10:                 for(i = 0; i < 8; i++)\
yading@10:                     OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\
yading@10:                 dst += stride;\
yading@10:                 tptr += 11;\
yading@10:             }\
yading@10: \
yading@10:             return;\
yading@10:         }\
yading@10:         else { /* No horizontal filter, output 8 lines to dst */\
yading@10:             r = 1-rnd;\
yading@10: \
yading@10:             for(j = 0; j < 8; j++) {\
yading@10:                 for(i = 0; i < 8; i++)\
yading@10:                     OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\
yading@10:                 src += stride;\
yading@10:                 dst += stride;\
yading@10:             }\
yading@10:             return;\
yading@10:         }\
yading@10:     }\
yading@10: \
yading@10:     /* Horizontal mode with no vertical mode */\
yading@10:     for(j = 0; j < 8; j++) {\
yading@10:         for(i = 0; i < 8; i++)\
yading@10:             OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\
yading@10:         dst += stride;\
yading@10:         src += stride;\
yading@10:     }\
yading@10: }\
yading@10: static void OPNAME ## pixels8x8_c(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int rnd){\
yading@10:     int i;\
yading@10:     for(i=0; i<8; i++){\
yading@10:         OP4(*(uint32_t*)(block  ), AV_RN32(pixels  ));\
yading@10:         OP4(*(uint32_t*)(block+4), AV_RN32(pixels+4));\
yading@10:         pixels+=line_size;\
yading@10:         block +=line_size;\
yading@10:     }\
yading@10: }
yading@10: 
yading@10: #define op_put(a, b) a = av_clip_uint8(b)
yading@10: #define op_avg(a, b) a = (a + av_clip_uint8(b) + 1) >> 1
yading@10: #define op4_avg(a, b) a = rnd_avg32(a, b)
yading@10: #define op4_put(a, b) a = b
yading@10: 
yading@10: VC1_MSPEL_MC(op_put, op4_put, put_)
yading@10: VC1_MSPEL_MC(op_avg, op4_avg, avg_)
yading@10: 
yading@10: /* pixel functions - really are entry points to vc1_mspel_mc */
yading@10: 
yading@10: #define PUT_VC1_MSPEL(a, b)\
yading@10: static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst,               \
yading@10:                                             const uint8_t *src,         \
yading@10:                                             ptrdiff_t stride, int rnd)  \
yading@10: {                                                                       \
yading@10:     put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                      \
yading@10: }                                                                       \
yading@10: static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst,               \
yading@10:                                             const uint8_t *src,         \
yading@10:                                             ptrdiff_t stride, int rnd)  \
yading@10: {                                                                       \
yading@10:     avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                      \
yading@10: }
yading@10: 
yading@10: PUT_VC1_MSPEL(1, 0)
yading@10: PUT_VC1_MSPEL(2, 0)
yading@10: PUT_VC1_MSPEL(3, 0)
yading@10: 
yading@10: PUT_VC1_MSPEL(0, 1)
yading@10: PUT_VC1_MSPEL(1, 1)
yading@10: PUT_VC1_MSPEL(2, 1)
yading@10: PUT_VC1_MSPEL(3, 1)
yading@10: 
yading@10: PUT_VC1_MSPEL(0, 2)
yading@10: PUT_VC1_MSPEL(1, 2)
yading@10: PUT_VC1_MSPEL(2, 2)
yading@10: PUT_VC1_MSPEL(3, 2)
yading@10: 
yading@10: PUT_VC1_MSPEL(0, 3)
yading@10: PUT_VC1_MSPEL(1, 3)
yading@10: PUT_VC1_MSPEL(2, 3)
yading@10: PUT_VC1_MSPEL(3, 3)
yading@10: 
yading@10: 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){
yading@10:     const int A=(8-x)*(8-y);
yading@10:     const int B=(  x)*(8-y);
yading@10:     const int C=(8-x)*(  y);
yading@10:     const int D=(  x)*(  y);
yading@10:     int i;
yading@10: 
yading@10:     av_assert2(x<8 && y<8 && x>=0 && y>=0);
yading@10: 
yading@10:     for(i=0; i<h; i++)
yading@10:     {
yading@10:         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;
yading@10:         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;
yading@10:         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;
yading@10:         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;
yading@10:         dst[4] = (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6;
yading@10:         dst[5] = (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6;
yading@10:         dst[6] = (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6;
yading@10:         dst[7] = (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6;
yading@10:         dst+= stride;
yading@10:         src+= stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void put_no_rnd_vc1_chroma_mc4_c(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y){
yading@10:     const int A=(8-x)*(8-y);
yading@10:     const int B=(  x)*(8-y);
yading@10:     const int C=(8-x)*(  y);
yading@10:     const int D=(  x)*(  y);
yading@10:     int i;
yading@10: 
yading@10:     av_assert2(x<8 && y<8 && x>=0 && y>=0);
yading@10: 
yading@10:     for(i=0; i<h; i++)
yading@10:     {
yading@10:         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;
yading@10:         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;
yading@10:         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;
yading@10:         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;
yading@10:         dst+= stride;
yading@10:         src+= stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: #define avg2(a,b) ((a+b+1)>>1)
yading@10: 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){
yading@10:     const int A=(8-x)*(8-y);
yading@10:     const int B=(  x)*(8-y);
yading@10:     const int C=(8-x)*(  y);
yading@10:     const int D=(  x)*(  y);
yading@10:     int i;
yading@10: 
yading@10:     av_assert2(x<8 && y<8 && x>=0 && y>=0);
yading@10: 
yading@10:     for(i=0; i<h; i++)
yading@10:     {
yading@10:         dst[0] = avg2(dst[0], ((A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6));
yading@10:         dst[1] = avg2(dst[1], ((A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6));
yading@10:         dst[2] = avg2(dst[2], ((A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6));
yading@10:         dst[3] = avg2(dst[3], ((A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6));
yading@10:         dst[4] = avg2(dst[4], ((A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6));
yading@10:         dst[5] = avg2(dst[5], ((A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6));
yading@10:         dst[6] = avg2(dst[6], ((A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6));
yading@10:         dst[7] = avg2(dst[7], ((A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6));
yading@10:         dst+= stride;
yading@10:         src+= stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
yading@10: 
yading@10: static void sprite_h_c(uint8_t *dst, const uint8_t *src, int offset, int advance, int count)
yading@10: {
yading@10:     while (count--) {
yading@10:         int a = src[(offset >> 16)    ];
yading@10:         int b = src[(offset >> 16) + 1];
yading@10:         *dst++ = a + ((b - a) * (offset&0xFFFF) >> 16);
yading@10:         offset += advance;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_always_inline void sprite_v_template(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
yading@10:                                             int two_sprites, const uint8_t *src2a, const uint8_t *src2b, int offset2,
yading@10:                                             int alpha, int scaled, int width)
yading@10: {
yading@10:     int a1, b1, a2, b2;
yading@10:     while (width--) {
yading@10:         a1 = *src1a++;
yading@10:         if (scaled) {
yading@10:             b1 = *src1b++;
yading@10:             a1 = a1 + ((b1 - a1) * offset1 >> 16);
yading@10:         }
yading@10:         if (two_sprites) {
yading@10:             a2 = *src2a++;
yading@10:             if (scaled > 1) {
yading@10:                 b2 = *src2b++;
yading@10:                 a2 = a2 + ((b2 - a2) * offset2 >> 16);
yading@10:             }
yading@10:             a1 = a1 + ((a2 - a1) * alpha >> 16);
yading@10:         }
yading@10:         *dst++ = a1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void sprite_v_single_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset, int width)
yading@10: {
yading@10:     sprite_v_template(dst, src1a, src1b, offset, 0, NULL, NULL, 0, 0, 1, width);
yading@10: }
yading@10: 
yading@10: static void sprite_v_double_noscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src2a, int alpha, int width)
yading@10: {
yading@10:     sprite_v_template(dst, src1a, NULL, 0, 1, src2a, NULL, 0, alpha, 0, width);
yading@10: }
yading@10: 
yading@10: static void sprite_v_double_onescale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
yading@10:                                                      const uint8_t *src2a, int alpha, int width)
yading@10: {
yading@10:     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, NULL, 0, alpha, 1, width);
yading@10: }
yading@10: 
yading@10: static void sprite_v_double_twoscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,
yading@10:                                                      const uint8_t *src2a, const uint8_t *src2b, int offset2,
yading@10:                                        int alpha, int width)
yading@10: {
yading@10:     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, src2b, offset2, alpha, 2, width);
yading@10: }
yading@10: 
yading@10: #endif
yading@10: 
yading@10: av_cold void ff_vc1dsp_init(VC1DSPContext* dsp) {
yading@10:     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
yading@10:     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
yading@10:     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
yading@10:     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
yading@10:     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;
yading@10:     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;
yading@10:     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;
yading@10:     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;
yading@10:     dsp->vc1_h_overlap = vc1_h_overlap_c;
yading@10:     dsp->vc1_v_overlap = vc1_v_overlap_c;
yading@10:     dsp->vc1_h_s_overlap = vc1_h_s_overlap_c;
yading@10:     dsp->vc1_v_s_overlap = vc1_v_s_overlap_c;
yading@10:     dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
yading@10:     dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
yading@10:     dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;
yading@10:     dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;
yading@10:     dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;
yading@10:     dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;
yading@10: 
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 0] = put_pixels8x8_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
yading@10:     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
yading@10: 
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 0] = avg_pixels8x8_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;
yading@10:     dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;
yading@10: 
yading@10:     dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= put_no_rnd_vc1_chroma_mc8_c;
yading@10:     dsp->avg_no_rnd_vc1_chroma_pixels_tab[0]= avg_no_rnd_vc1_chroma_mc8_c;
yading@10:     dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = put_no_rnd_vc1_chroma_mc4_c;
yading@10: 
yading@10: #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
yading@10:     dsp->sprite_h = sprite_h_c;
yading@10:     dsp->sprite_v_single = sprite_v_single_c;
yading@10:     dsp->sprite_v_double_noscale = sprite_v_double_noscale_c;
yading@10:     dsp->sprite_v_double_onescale = sprite_v_double_onescale_c;
yading@10:     dsp->sprite_v_double_twoscale = sprite_v_double_twoscale_c;
yading@10: #endif
yading@10: 
yading@10:     if (HAVE_ALTIVEC)
yading@10:         ff_vc1dsp_init_altivec(dsp);
yading@10:     if (ARCH_X86)
yading@10:         ff_vc1dsp_init_x86(dsp);
yading@10: }