Add DXVA2 HEVC HWAccel
[ffmpeg.git] / libavcodec / vc1dsp.c
index 47b69c8..a193dd7 100644 (file)
  * VC-1 and WMV3 decoder - DSP functions
  * Copyright (c) 2006 Konstantin Shishkov
  *
- * This file is part of FFmpeg.
+ * This file is part of Libav.
  *
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
  * version 2.1 of the License, or (at your option) any later version.
  *
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
 /**
-* @file
+ * @file
  * VC-1 and WMV3 decoder
  *
  */
 
-#include "dsputil.h"
+#include "libavutil/common.h"
+#include "h264chroma.h"
+#include "qpeldsp.h"
+#include "vc1dsp.h"
+#include "startcode.h"
 
-
-/** Apply overlap transform to horizontal edge
-*/
-static void vc1_v_overlap_c(uint8_t* src, int stride)
+/* Apply overlap transform to horizontal edge */
+static void vc1_v_overlap_c(uint8_t *src, int stride)
 {
     int i;
     int a, b, c, d;
     int d1, d2;
     int rnd = 1;
-    for(i = 0; i < 8; i++) {
-        a = src[-2*stride];
-        b = src[-stride];
-        c = src[0];
-        d = src[stride];
+    for (i = 0; i < 8; i++) {
+        a  = src[-2 * stride];
+        b  = src[-stride];
+        c  = src[0];
+        d  = src[stride];
         d1 = (a - d + 3 + rnd) >> 3;
         d2 = (a - d + b - c + 4 - rnd) >> 3;
 
-        src[-2*stride] = a - d1;
-        src[-stride] = av_clip_uint8(b - d2);
-        src[0] = av_clip_uint8(c + d2);
-        src[stride] = d + d1;
+        src[-2 * stride] = a - d1;
+        src[-stride]     = av_clip_uint8(b - d2);
+        src[0]           = av_clip_uint8(c + d2);
+        src[stride]      = d + d1;
         src++;
         rnd = !rnd;
     }
 }
 
-/** Apply overlap transform to vertical edge
-*/
-static void vc1_h_overlap_c(uint8_t* src, int stride)
+/* Apply overlap transform to vertical edge */
+static void vc1_h_overlap_c(uint8_t *src, int stride)
 {
     int i;
     int a, b, c, d;
     int d1, d2;
     int rnd = 1;
-    for(i = 0; i < 8; i++) {
-        a = src[-2];
-        b = src[-1];
-        c = src[0];
-        d = src[1];
+    for (i = 0; i < 8; i++) {
+        a  = src[-2];
+        b  = src[-1];
+        c  = src[0];
+        d  = src[1];
         d1 = (a - d + 3 + rnd) >> 3;
         d2 = (a - d + b - c + 4 - rnd) >> 3;
 
         src[-2] = a - d1;
         src[-1] = av_clip_uint8(b - d2);
-        src[0] = av_clip_uint8(c + d2);
-        src[1] = d + d1;
-        src += stride;
-        rnd = !rnd;
+        src[0]  = av_clip_uint8(c + d2);
+        src[1]  = d + d1;
+        src    += stride;
+        rnd     = !rnd;
+    }
+}
+
+static void vc1_v_s_overlap_c(int16_t *top, int16_t *bottom)
+{
+    int i;
+    int a, b, c, d;
+    int d1, d2;
+    int rnd1 = 4, rnd2 = 3;
+    for (i = 0; i < 8; i++) {
+        a  = top[48];
+        b  = top[56];
+        c  = bottom[0];
+        d  = bottom[8];
+        d1 = a - d;
+        d2 = a - d + b - c;
+
+        top[48]   = ((a << 3) - d1 + rnd1) >> 3;
+        top[56]   = ((b << 3) - d2 + rnd2) >> 3;
+        bottom[0] = ((c << 3) + d2 + rnd1) >> 3;
+        bottom[8] = ((d << 3) + d1 + rnd2) >> 3;
+
+        bottom++;
+        top++;
+        rnd2 = 7 - rnd2;
+        rnd1 = 7 - rnd1;
+    }
+}
+
+static void vc1_h_s_overlap_c(int16_t *left, int16_t *right)
+{
+    int i;
+    int a, b, c, d;
+    int d1, d2;
+    int rnd1 = 4, rnd2 = 3;
+    for (i = 0; i < 8; i++) {
+        a  = left[6];
+        b  = left[7];
+        c  = right[0];
+        d  = right[1];
+        d1 = a - d;
+        d2 = a - d + b - c;
+
+        left[6]  = ((a << 3) - d1 + rnd1) >> 3;
+        left[7]  = ((b << 3) - d2 + rnd2) >> 3;
+        right[0] = ((c << 3) + d2 + rnd1) >> 3;
+        right[1] = ((d << 3) + d1 + rnd2) >> 3;
+
+        right += 8;
+        left  += 8;
+        rnd2   = 7 - rnd2;
+        rnd1   = 7 - rnd1;
     }
 }
 
@@ -86,33 +139,38 @@ static void vc1_h_overlap_c(uint8_t* src, int stride)
  * @return whether other 3 pairs should be filtered or not
  * @see 8.6
  */
-static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
-    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
-
-    int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
+static av_always_inline int vc1_filter_line(uint8_t *src, int stride, int pq)
+{
+    int a0 = (2 * (src[-2 * stride] - src[1 * stride]) -
+              5 * (src[-1 * stride] - src[0 * stride]) + 4) >> 3;
     int a0_sign = a0 >> 31;        /* Store sign */
+
     a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
-    if(a0 < pq){
-        int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
-        int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
-        if(a1 < a0 || a2 < a0){
-            int clip = src[-1*stride] - src[ 0*stride];
+    if (a0 < pq) {
+        int a1 = FFABS((2 * (src[-4 * stride] - src[-1 * stride]) -
+                        5 * (src[-3 * stride] - src[-2 * stride]) + 4) >> 3);
+        int a2 = FFABS((2 * (src[ 0 * stride] - src[ 3 * stride]) -
+                        5 * (src[ 1 * stride] - src[ 2 * stride]) + 4) >> 3);
+        if (a1 < a0 || a2 < a0) {
+            int clip      = src[-1 * stride] - src[0 * stride];
             int clip_sign = clip >> 31;
-            clip = ((clip ^ clip_sign) - clip_sign)>>1;
-            if(clip){
-                int a3 = FFMIN(a1, a2);
-                int d = 5 * (a3 - a0);
+
+            clip = ((clip ^ clip_sign) - clip_sign) >> 1;
+            if (clip) {
+                int a3     = FFMIN(a1, a2);
+                int d      = 5 * (a3 - a0);
                 int d_sign = (d >> 31);
-                d = ((d ^ d_sign) - d_sign) >> 3;
+
+                d       = ((d ^ d_sign) - d_sign) >> 3;
                 d_sign ^= a0_sign;
 
-                if( d_sign ^ clip_sign )
+                if (d_sign ^ clip_sign)
                     d = 0;
-                else{
+                else {
                     d = FFMIN(d, clip);
-                    d = (d ^ d_sign) - d_sign;          /* Restore sign */
-                    src[-1*stride] = cm[src[-1*stride] - d];
-                    src[ 0*stride] = cm[src[ 0*stride] + d];
+                    d = (d ^ d_sign) - d_sign; /* Restore sign */
+                    src[-1 * stride] = av_clip_uint8(src[-1 * stride] - d);
+                    src[ 0 * stride] = av_clip_uint8(src[ 0 * stride] + d);
                 }
                 return 1;
             }
@@ -130,17 +188,18 @@ static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
  * @param pq block quantizer
  * @see 8.6
  */
-static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
+static inline void vc1_loop_filter(uint8_t *src, int step, int stride,
+                                   int len, int pq)
 {
     int i;
     int filt3;
 
-    for(i = 0; i < len; i += 4){
-        filt3 = vc1_filter_line(src + 2*step, stride, pq);
-        if(filt3){
-            vc1_filter_line(src + 0*step, stride, pq);
-            vc1_filter_line(src + 1*step, stride, pq);
-            vc1_filter_line(src + 3*step, stride, pq);
+    for (i = 0; i < len; i += 4) {
+        filt3 = vc1_filter_line(src + 2 * step, stride, pq);
+        if (filt3) {
+            vc1_filter_line(src + 0 * step, stride, pq);
+            vc1_filter_line(src + 1 * step, stride, pq);
+            vc1_filter_line(src + 3 * step, stride, pq);
         }
         src += step * 4;
     }
@@ -176,51 +235,51 @@ static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)
     vc1_loop_filter(src, stride, 1, 16, pq);
 }
 
-/** Do inverse transform on 8x8 block
-*/
-static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
+/* Do inverse transform on 8x8 block */
+static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
     int dc = block[0];
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+
     dc = (3 * dc +  1) >> 1;
     dc = (3 * dc + 16) >> 5;
-    for(i = 0; i < 8; i++){
-        dest[0] = cm[dest[0]+dc];
-        dest[1] = cm[dest[1]+dc];
-        dest[2] = cm[dest[2]+dc];
-        dest[3] = cm[dest[3]+dc];
-        dest[4] = cm[dest[4]+dc];
-        dest[5] = cm[dest[5]+dc];
-        dest[6] = cm[dest[6]+dc];
-        dest[7] = cm[dest[7]+dc];
+
+    for (i = 0; i < 8; i++) {
+        dest[0] = av_clip_uint8(dest[0] + dc);
+        dest[1] = av_clip_uint8(dest[1] + dc);
+        dest[2] = av_clip_uint8(dest[2] + dc);
+        dest[3] = av_clip_uint8(dest[3] + dc);
+        dest[4] = av_clip_uint8(dest[4] + dc);
+        dest[5] = av_clip_uint8(dest[5] + dc);
+        dest[6] = av_clip_uint8(dest[6] + dc);
+        dest[7] = av_clip_uint8(dest[7] + dc);
         dest += linesize;
     }
 }
 
-static void vc1_inv_trans_8x8_c(DCTELEM block[64])
+static void vc1_inv_trans_8x8_c(int16_t block[64])
 {
     int i;
-    register int t1,t2,t3,t4,t5,t6,t7,t8;
-    DCTELEM *src, *dst;
+    register int t1, t2, t3, t4, t5, t6, t7, t8;
+    int16_t *src, *dst, temp[64];
 
     src = block;
-    dst = block;
-    for(i = 0; i < 8; i++){
-        t1 = 12 * (src[0] + src[4]) + 4;
-        t2 = 12 * (src[0] - src[4]) + 4;
-        t3 = 16 * src[2] +  6 * src[6];
-        t4 =  6 * src[2] - 16 * src[6];
+    dst = temp;
+    for (i = 0; i < 8; i++) {
+        t1 = 12 * (src[ 0] + src[32]) + 4;
+        t2 = 12 * (src[ 0] - src[32]) + 4;
+        t3 = 16 * src[16] +  6 * src[48];
+        t4 =  6 * src[16] - 16 * src[48];
 
         t5 = t1 + t3;
         t6 = t2 + t4;
         t7 = t2 - t4;
         t8 = t1 - t3;
 
-        t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
-        t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
-        t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
-        t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
+        t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
+        t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
+        t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
+        t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 
         dst[0] = (t5 + t1) >> 3;
         dst[1] = (t6 + t2) >> 3;
@@ -231,13 +290,13 @@ static void vc1_inv_trans_8x8_c(DCTELEM block[64])
         dst[6] = (t6 - t2) >> 3;
         dst[7] = (t5 - t1) >> 3;
 
-        src += 8;
+        src += 1;
         dst += 8;
     }
 
-    src = block;
+    src = temp;
     dst = block;
-    for(i = 0; i < 8; i++){
+    for (i = 0; i < 8; i++) {
         t1 = 12 * (src[ 0] + src[32]) + 64;
         t2 = 12 * (src[ 0] - src[32]) + 64;
         t3 = 16 * src[16] +  6 * src[48];
@@ -267,38 +326,38 @@ static void vc1_inv_trans_8x8_c(DCTELEM block[64])
     }
 }
 
-/** Do inverse transform on 8x4 part of block
-*/
-static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
+/* Do inverse transform on 8x4 part of block */
+static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
     int dc = block[0];
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
-    dc = 3 * dc +  1) >> 1;
+
+    dc =  (3 * dc +  1) >> 1;
     dc = (17 * dc + 64) >> 7;
-    for(i = 0; i < 4; i++){
-        dest[0] = cm[dest[0]+dc];
-        dest[1] = cm[dest[1]+dc];
-        dest[2] = cm[dest[2]+dc];
-        dest[3] = cm[dest[3]+dc];
-        dest[4] = cm[dest[4]+dc];
-        dest[5] = cm[dest[5]+dc];
-        dest[6] = cm[dest[6]+dc];
-        dest[7] = cm[dest[7]+dc];
+
+    for (i = 0; i < 4; i++) {
+        dest[0] = av_clip_uint8(dest[0] + dc);
+        dest[1] = av_clip_uint8(dest[1] + dc);
+        dest[2] = av_clip_uint8(dest[2] + dc);
+        dest[3] = av_clip_uint8(dest[3] + dc);
+        dest[4] = av_clip_uint8(dest[4] + dc);
+        dest[5] = av_clip_uint8(dest[5] + dc);
+        dest[6] = av_clip_uint8(dest[6] + dc);
+        dest[7] = av_clip_uint8(dest[7] + dc);
         dest += linesize;
     }
 }
 
-static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
+static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
-    register int t1,t2,t3,t4,t5,t6,t7,t8;
-    DCTELEM *src, *dst;
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+    register int t1, t2, t3, t4, t5, t6, t7, t8;
+    int16_t *src, *dst;
 
     src = block;
     dst = block;
-    for(i = 0; i < 4; i++){
+
+    for (i = 0; i < 4; i++) {
         t1 = 12 * (src[0] + src[4]) + 4;
         t2 = 12 * (src[0] - src[4]) + 4;
         t3 = 16 * src[2] +  6 * src[6];
@@ -328,50 +387,50 @@ static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
     }
 
     src = block;
-    for(i = 0; i < 8; i++){
+    for (i = 0; i < 8; i++) {
         t1 = 17 * (src[ 0] + src[16]) + 64;
         t2 = 17 * (src[ 0] - src[16]) + 64;
         t3 = 22 * src[ 8] + 10 * src[24];
         t4 = 22 * src[24] - 10 * src[ 8];
 
-        dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)];
-        dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)];
-        dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)];
-        dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)];
+        dest[0 * linesize] = av_clip_uint8(dest[0 * linesize] + ((t1 + t3) >> 7));
+        dest[1 * linesize] = av_clip_uint8(dest[1 * linesize] + ((t2 - t4) >> 7));
+        dest[2 * linesize] = av_clip_uint8(dest[2 * linesize] + ((t2 + t4) >> 7));
+        dest[3 * linesize] = av_clip_uint8(dest[3 * linesize] + ((t1 - t3) >> 7));
 
-        src ++;
+        src++;
         dest++;
     }
 }
 
-/** Do inverse transform on 4x8 parts of block
-*/
-static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
+/* Do inverse transform on 4x8 parts of block */
+static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
     int dc = block[0];
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+
     dc = (17 * dc +  4) >> 3;
     dc = (12 * dc + 64) >> 7;
-    for(i = 0; i < 8; i++){
-        dest[0] = cm[dest[0]+dc];
-        dest[1] = cm[dest[1]+dc];
-        dest[2] = cm[dest[2]+dc];
-        dest[3] = cm[dest[3]+dc];
+
+    for (i = 0; i < 8; i++) {
+        dest[0] = av_clip_uint8(dest[0] + dc);
+        dest[1] = av_clip_uint8(dest[1] + dc);
+        dest[2] = av_clip_uint8(dest[2] + dc);
+        dest[3] = av_clip_uint8(dest[3] + dc);
         dest += linesize;
     }
 }
 
-static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
+static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
-    register int t1,t2,t3,t4,t5,t6,t7,t8;
-    DCTELEM *src, *dst;
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+    register int t1, t2, t3, t4, t5, t6, t7, t8;
+    int16_t *src, *dst;
 
     src = block;
     dst = block;
-    for(i = 0; i < 8; i++){
+
+    for (i = 0; i < 8; i++) {
         t1 = 17 * (src[0] + src[2]) + 4;
         t2 = 17 * (src[0] - src[2]) + 4;
         t3 = 22 * src[1] + 10 * src[3];
@@ -387,7 +446,7 @@ static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
     }
 
     src = block;
-    for(i = 0; i < 4; i++){
+    for (i = 0; i < 4; i++) {
         t1 = 12 * (src[ 0] + src[32]) + 64;
         t2 = 12 * (src[ 0] - src[32]) + 64;
         t3 = 16 * src[16] +  6 * src[48];
@@ -403,48 +462,47 @@ static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 
-        dest[0*linesize] = cm[dest[0*linesize] + ((t5 + t1) >> 7)];
-        dest[1*linesize] = cm[dest[1*linesize] + ((t6 + t2) >> 7)];
-        dest[2*linesize] = cm[dest[2*linesize] + ((t7 + t3) >> 7)];
-        dest[3*linesize] = cm[dest[3*linesize] + ((t8 + t4) >> 7)];
-        dest[4*linesize] = cm[dest[4*linesize] + ((t8 - t4 + 1) >> 7)];
-        dest[5*linesize] = cm[dest[5*linesize] + ((t7 - t3 + 1) >> 7)];
-        dest[6*linesize] = cm[dest[6*linesize] + ((t6 - t2 + 1) >> 7)];
-        dest[7*linesize] = cm[dest[7*linesize] + ((t5 - t1 + 1) >> 7)];
+        dest[0 * linesize] = av_clip_uint8(dest[0 * linesize] + ((t5 + t1)     >> 7));
+        dest[1 * linesize] = av_clip_uint8(dest[1 * linesize] + ((t6 + t2)     >> 7));
+        dest[2 * linesize] = av_clip_uint8(dest[2 * linesize] + ((t7 + t3)     >> 7));
+        dest[3 * linesize] = av_clip_uint8(dest[3 * linesize] + ((t8 + t4)     >> 7));
+        dest[4 * linesize] = av_clip_uint8(dest[4 * linesize] + ((t8 - t4 + 1) >> 7));
+        dest[5 * linesize] = av_clip_uint8(dest[5 * linesize] + ((t7 - t3 + 1) >> 7));
+        dest[6 * linesize] = av_clip_uint8(dest[6 * linesize] + ((t6 - t2 + 1) >> 7));
+        dest[7 * linesize] = av_clip_uint8(dest[7 * linesize] + ((t5 - t1 + 1) >> 7));
 
-        src ++;
+        src++;
         dest++;
     }
 }
 
-/** Do inverse transform on 4x4 part of block
-*/
-static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
+/* Do inverse transform on 4x4 part of block */
+static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
     int dc = block[0];
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+
     dc = (17 * dc +  4) >> 3;
     dc = (17 * dc + 64) >> 7;
-    for(i = 0; i < 4; i++){
-        dest[0] = cm[dest[0]+dc];
-        dest[1] = cm[dest[1]+dc];
-        dest[2] = cm[dest[2]+dc];
-        dest[3] = cm[dest[3]+dc];
+
+    for (i = 0; i < 4; i++) {
+        dest[0] = av_clip_uint8(dest[0] + dc);
+        dest[1] = av_clip_uint8(dest[1] + dc);
+        dest[2] = av_clip_uint8(dest[2] + dc);
+        dest[3] = av_clip_uint8(dest[3] + dc);
         dest += linesize;
     }
 }
 
-static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
+static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, int16_t *block)
 {
     int i;
-    register int t1,t2,t3,t4;
-    DCTELEM *src, *dst;
-    const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
+    register int t1, t2, t3, t4;
+    int16_t *src, *dst;
 
     src = block;
     dst = block;
-    for(i = 0; i < 4; i++){
+    for (i = 0; i < 4; i++) {
         t1 = 17 * (src[0] + src[2]) + 4;
         t2 = 17 * (src[0] - src[2]) + 4;
         t3 = 22 * src[1] + 10 * src[3];
@@ -460,117 +518,128 @@ static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
     }
 
     src = block;
-    for(i = 0; i < 4; i++){
-        t1 = 17 * (src[ 0] + src[16]) + 64;
-        t2 = 17 * (src[ 0] - src[16]) + 64;
-        t3 = 22 * src[ 8] + 10 * src[24];
-        t4 = 22 * src[24] - 10 * src[ 8];
+    for (i = 0; i < 4; i++) {
+        t1 = 17 * (src[0] + src[16]) + 64;
+        t2 = 17 * (src[0] - src[16]) + 64;
+        t3 = 22 * src[8] + 10 * src[24];
+        t4 = 22 * src[24] - 10 * src[8];
 
-        dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)];
-        dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)];
-        dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)];
-        dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)];
+        dest[0 * linesize] = av_clip_uint8(dest[0 * linesize] + ((t1 + t3) >> 7));
+        dest[1 * linesize] = av_clip_uint8(dest[1 * linesize] + ((t2 - t4) >> 7));
+        dest[2 * linesize] = av_clip_uint8(dest[2 * linesize] + ((t2 + t4) >> 7));
+        dest[3 * linesize] = av_clip_uint8(dest[3 * linesize] + ((t1 - t3) >> 7));
 
-        src ++;
+        src++;
         dest++;
     }
 }
 
 /* motion compensation functions */
-/** Filter in case of 2 filters */
-#define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \
-static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \
-{                                                                       \
-    switch(mode){                                                       \
-    case 0: /* no shift - should not occur */                           \
-        return 0;                                                       \
-    case 1: /* 1/4 shift */                                             \
-        return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \
-    case 2: /* 1/2 shift */                                             \
-        return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \
-    case 3: /* 3/4 shift */                                             \
-        return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \
-    }                                                                   \
-    return 0; /* should not occur */                                    \
-}
-
-VC1_MSPEL_FILTER_16B(ver, uint8_t);
-VC1_MSPEL_FILTER_16B(hor, int16_t);
-
-
-/** Filter used to interpolate fractional pel values
- */
-static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
+
+/* Filter in case of 2 filters */
+#define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                       \
+static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, \
+                                                                int stride,   \
+                                                                int mode)     \
+{                                                                             \
+    switch(mode) {                                                            \
+    case 0: /* no shift - should not occur */                                 \
+        return 0;                                                             \
+    case 1: /* 1/4 shift */                                                   \
+        return -4 * src[-stride] + 53 * src[0] +                              \
+               18 * src[stride]  -  3 * src[stride * 2];                      \
+    case 2: /* 1/2 shift */                                                   \
+        return -1 * src[-stride] +  9 * src[0] +                              \
+                9 * src[stride]  -  1 * src[stride * 2];                      \
+    case 3: /* 3/4 shift */                                                   \
+        return -3 * src[-stride] + 18 * src[0] +                              \
+               53 * src[stride]  -  4 * src[stride * 2];                      \
+    }                                                                         \
+    return 0; /* should not occur */                                          \
+}
+
+VC1_MSPEL_FILTER_16B(ver, uint8_t)
+VC1_MSPEL_FILTER_16B(hor, int16_t)
+
+/* Filter used to interpolate fractional pel values */
+static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride,
+                                             int mode, int r)
 {
-    switch(mode){
-    case 0: //no shift
+    switch (mode) {
+    case 0: // no shift
         return src[0];
     case 1: // 1/4 shift
-        return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
+        return (-4 * src[-stride] + 53 * src[0] +
+                18 * src[stride]  -  3 * src[stride * 2] + 32 - r) >> 6;
     case 2: // 1/2 shift
-        return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
+        return (-1 * src[-stride] +  9 * src[0] +
+                 9 * src[stride]  -  1 * src[stride * 2] + 8 - r) >> 4;
     case 3: // 3/4 shift
-        return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
+        return (-3 * src[-stride] + 18 * src[0] +
+                53 * src[stride]  -  4 * src[stride * 2] + 32 - r) >> 6;
     }
-    return 0; //should not occur
+    return 0; // should not occur
 }
 
-/** Function used to do motion compensation with bicubic interpolation
- */
-#define VC1_MSPEL_MC(OP, OPNAME)\
-static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
-{\
-    int     i, j;\
-\
-    if (vmode) { /* Horizontal filter to apply */\
-        int r;\
-\
-        if (hmode) { /* Vertical filter to apply, output to tmp */\
-            static const int shift_value[] = { 0, 5, 1, 5 };\
-            int              shift = (shift_value[hmode]+shift_value[vmode])>>1;\
-            int16_t          tmp[11*8], *tptr = tmp;\
-\
-            r = (1<<(shift-1)) + rnd-1;\
-\
-            src -= 1;\
-            for(j = 0; j < 8; j++) {\
-                for(i = 0; i < 11; i++)\
-                    tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\
-                src += stride;\
-                tptr += 11;\
-            }\
-\
-            r = 64-rnd;\
-            tptr = tmp+1;\
-            for(j = 0; j < 8; j++) {\
-                for(i = 0; i < 8; i++)\
-                    OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\
-                dst += stride;\
-                tptr += 11;\
-            }\
-\
-            return;\
-        }\
-        else { /* No horizontal filter, output 8 lines to dst */\
-            r = 1-rnd;\
-\
-            for(j = 0; j < 8; j++) {\
-                for(i = 0; i < 8; i++)\
-                    OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\
-                src += stride;\
-                dst += stride;\
-            }\
-            return;\
-        }\
-    }\
-\
-    /* Horizontal mode with no vertical mode */\
-    for(j = 0; j < 8; j++) {\
-        for(i = 0; i < 8; i++)\
-            OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\
-        dst += stride;\
-        src += stride;\
-    }\
+/* Function used to do motion compensation with bicubic interpolation */
+#define VC1_MSPEL_MC(OP, OPNAME)                                              \
+static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst,             \
+                                                    const uint8_t *src,       \
+                                                    int stride,               \
+                                                    int hmode,                \
+                                                    int vmode,                \
+                                                    int rnd)                  \
+{                                                                             \
+    int i, j;                                                                 \
+                                                                              \
+    if (vmode) { /* Horizontal filter to apply */                             \
+        int r;                                                                \
+                                                                              \
+        if (hmode) { /* Vertical filter to apply, output to tmp */            \
+            static const int shift_value[] = { 0, 5, 1, 5 };                  \
+            int shift = (shift_value[hmode] + shift_value[vmode]) >> 1;       \
+            int16_t tmp[11 * 8], *tptr = tmp;                                 \
+                                                                              \
+            r = (1 << (shift - 1)) + rnd - 1;                                 \
+                                                                              \
+            src -= 1;                                                         \
+            for (j = 0; j < 8; j++) {                                         \
+                for (i = 0; i < 11; i++)                                      \
+                    tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode) + r) >> shift; \
+                src  += stride;                                               \
+                tptr += 11;                                                   \
+            }                                                                 \
+                                                                              \
+            r    = 64 - rnd;                                                  \
+            tptr = tmp + 1;                                                   \
+            for (j = 0; j < 8; j++) {                                         \
+                for (i = 0; i < 8; i++)                                       \
+                    OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode) + r) >> 7); \
+                dst  += stride;                                               \
+                tptr += 11;                                                   \
+            }                                                                 \
+                                                                              \
+            return;                                                           \
+        } else { /* No horizontal filter, output 8 lines to dst */            \
+            r = 1 - rnd;                                                      \
+                                                                              \
+            for (j = 0; j < 8; j++) {                                         \
+                for (i = 0; i < 8; i++)                                       \
+                    OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));  \
+                src += stride;                                                \
+                dst += stride;                                                \
+            }                                                                 \
+            return;                                                           \
+        }                                                                     \
+    }                                                                         \
+                                                                              \
+    /* Horizontal mode with no vertical mode */                               \
+    for (j = 0; j < 8; j++) {                                                 \
+        for (i = 0; i < 8; i++)                                               \
+            OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));             \
+        dst += stride;                                                        \
+        src += stride;                                                        \
+    }                                                                         \
 }
 
 #define op_put(a, b) a = av_clip_uint8(b)
@@ -581,12 +650,18 @@ VC1_MSPEL_MC(op_avg, avg_)
 
 /* pixel functions - really are entry points to vc1_mspel_mc */
 
-#define PUT_VC1_MSPEL(a, b)\
-static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
-     put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
-}\
-static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
-     avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
+#define PUT_VC1_MSPEL(a, b)                                                   \
+static void put_vc1_mspel_mc ## a ## b ## _c(uint8_t *dst,                    \
+                                             const uint8_t *src,              \
+                                             ptrdiff_t stride, int rnd)       \
+{                                                                             \
+    put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                            \
+}                                                                             \
+static void avg_vc1_mspel_mc ## a ## b ## _c(uint8_t *dst,                    \
+                                             const uint8_t *src,              \
+                                             ptrdiff_t stride, int rnd)       \
+{                                                                             \
+    avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                            \
 }
 
 PUT_VC1_MSPEL(1, 0)
@@ -608,34 +683,235 @@ PUT_VC1_MSPEL(1, 3)
 PUT_VC1_MSPEL(2, 3)
 PUT_VC1_MSPEL(3, 3)
 
-av_cold void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
-    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
-    dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
-    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
-    dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
+
+static void put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src,
+                                 ptrdiff_t stride, int rnd)
+{
+    ff_put_pixels8x8_c(dst, src, stride);
+}
+
+static void avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src,
+                                 ptrdiff_t stride, int rnd)
+{
+    ff_avg_pixels8x8_c(dst, src, stride);
+}
+
+#define chroma_mc(a) \
+    ((A * src[a] + B * src[a + 1] + \
+      C * src[stride + a] + D * src[stride + a + 1] + 32 - 4) >> 6)
+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)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B =     (x) * (8 - y);
+    const int C = (8 - x) *     (y);
+    const int D =     (x) *     (y);
+    int i;
+
+    assert(x < 8 && y < 8 && x >= 0 && y >= 0);
+
+    for (i = 0; i < h; i++) {
+        dst[0] = chroma_mc(0);
+        dst[1] = chroma_mc(1);
+        dst[2] = chroma_mc(2);
+        dst[3] = chroma_mc(3);
+        dst[4] = chroma_mc(4);
+        dst[5] = chroma_mc(5);
+        dst[6] = chroma_mc(6);
+        dst[7] = chroma_mc(7);
+        dst += stride;
+        src += stride;
+    }
+}
+
+static void put_no_rnd_vc1_chroma_mc4_c(uint8_t *dst, uint8_t *src,
+                                        int stride, int h, int x, int y)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B =     (x) * (8 - y);
+    const int C = (8 - x) *     (y);
+    const int D =     (x) *     (y);
+    int i;
+
+    assert(x < 8 && y < 8 && x >= 0 && y >= 0);
+
+    for (i = 0; i < h; i++) {
+        dst[0] = chroma_mc(0);
+        dst[1] = chroma_mc(1);
+        dst[2] = chroma_mc(2);
+        dst[3] = chroma_mc(3);
+        dst += stride;
+        src += stride;
+    }
+}
+
+#define avg2(a, b) (((a) + (b) + 1) >> 1)
+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)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B =     (x) * (8 - y);
+    const int C = (8 - x) *     (y);
+    const int D =     (x) *     (y);
+    int i;
+
+    assert(x < 8 && y < 8 && x >= 0 && y >= 0);
+
+    for (i = 0; i < h; i++) {
+        dst[0] = avg2(dst[0], chroma_mc(0));
+        dst[1] = avg2(dst[1], chroma_mc(1));
+        dst[2] = avg2(dst[2], chroma_mc(2));
+        dst[3] = avg2(dst[3], chroma_mc(3));
+        dst[4] = avg2(dst[4], chroma_mc(4));
+        dst[5] = avg2(dst[5], chroma_mc(5));
+        dst[6] = avg2(dst[6], chroma_mc(6));
+        dst[7] = avg2(dst[7], chroma_mc(7));
+        dst += stride;
+        src += stride;
+    }
+}
+
+static void avg_no_rnd_vc1_chroma_mc4_c(uint8_t *dst /* align 8 */,
+                                        uint8_t *src /* align 1 */,
+                                        int stride, int h, int x, int y)
+{
+    const int A = (8 - x) * (8 - y);
+    const int B = (    x) * (8 - y);
+    const int C = (8 - x) * (    y);
+    const int D = (    x) * (    y);
+    int i;
+
+    assert(x < 8 && y < 8 && x >= 0 && y >= 0);
+
+    for (i = 0; i < h; i++) {
+        dst[0] = avg2(dst[0], chroma_mc(0));
+        dst[1] = avg2(dst[1], chroma_mc(1));
+        dst[2] = avg2(dst[2], chroma_mc(2));
+        dst[3] = avg2(dst[3], chroma_mc(3));
+        dst += stride;
+        src += stride;
+    }
+}
+
+#if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
+
+static void sprite_h_c(uint8_t *dst, const uint8_t *src, int offset,
+                       int advance, int count)
+{
+    while (count--) {
+        int a = src[(offset >> 16)];
+        int b = src[(offset >> 16) + 1];
+        *dst++  = a + ((b - a) * (offset & 0xFFFF) >> 16);
+        offset += advance;
+    }
+}
+
+static av_always_inline void sprite_v_template(uint8_t *dst,
+                                               const uint8_t *src1a,
+                                               const uint8_t *src1b,
+                                               int offset1,
+                                               int two_sprites,
+                                               const uint8_t *src2a,
+                                               const uint8_t *src2b,
+                                               int offset2,
+                                               int alpha, int scaled,
+                                               int width)
+{
+    int a1, b1, a2, b2;
+    while (width--) {
+        a1 = *src1a++;
+        if (scaled) {
+            b1 = *src1b++;
+            a1 = a1 + ((b1 - a1) * offset1 >> 16);
+        }
+        if (two_sprites) {
+            a2 = *src2a++;
+            if (scaled > 1) {
+                b2 = *src2b++;
+                a2 = a2 + ((b2 - a2) * offset2 >> 16);
+            }
+            a1 = a1 + ((a2 - a1) * alpha >> 16);
+        }
+        *dst++ = a1;
+    }
+}
+
+static void sprite_v_single_c(uint8_t *dst, const uint8_t *src1a,
+                              const uint8_t *src1b,
+                              int offset, int width)
+{
+    sprite_v_template(dst, src1a, src1b, offset, 0, NULL, NULL, 0, 0, 1, width);
+}
+
+static void sprite_v_double_noscale_c(uint8_t *dst, const uint8_t *src1a,
+                                      const uint8_t *src2a,
+                                      int alpha, int width)
+{
+    sprite_v_template(dst, src1a, NULL, 0, 1, src2a, NULL, 0, alpha, 0, width);
+}
+
+static void sprite_v_double_onescale_c(uint8_t *dst,
+                                       const uint8_t *src1a,
+                                       const uint8_t *src1b,
+                                       int offset1,
+                                       const uint8_t *src2a,
+                                       int alpha, int width)
+{
+    sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, NULL, 0, alpha, 1,
+                      width);
+}
+
+static void sprite_v_double_twoscale_c(uint8_t *dst,
+                                       const uint8_t *src1a,
+                                       const uint8_t *src1b,
+                                       int offset1,
+                                       const uint8_t *src2a,
+                                       const uint8_t *src2b,
+                                       int offset2,
+                                       int alpha,
+                                       int width)
+{
+    sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, src2b, offset2,
+                      alpha, 2, width);
+}
+
+#endif /* CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER */
+
+av_cold void ff_vc1dsp_init(VC1DSPContext *dsp)
+{
+    dsp->vc1_inv_trans_8x8    = vc1_inv_trans_8x8_c;
+    dsp->vc1_inv_trans_4x8    = vc1_inv_trans_4x8_c;
+    dsp->vc1_inv_trans_8x4    = vc1_inv_trans_8x4_c;
+    dsp->vc1_inv_trans_4x4    = vc1_inv_trans_4x4_c;
     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;
     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;
     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;
     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;
-    dsp->vc1_h_overlap = vc1_h_overlap_c;
-    dsp->vc1_v_overlap = vc1_v_overlap_c;
-    dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
-    dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
-    dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;
-    dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;
-    dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;
-    dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;
-
-    dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
-    dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
-    dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
-    dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
-    dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
-    dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
-    dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
-    dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
-    dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
-    dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
+
+    dsp->vc1_h_overlap        = vc1_h_overlap_c;
+    dsp->vc1_v_overlap        = vc1_v_overlap_c;
+    dsp->vc1_h_s_overlap      = vc1_h_s_overlap_c;
+    dsp->vc1_v_s_overlap      = vc1_v_s_overlap_c;
+
+    dsp->vc1_v_loop_filter4   = vc1_v_loop_filter4_c;
+    dsp->vc1_h_loop_filter4   = vc1_h_loop_filter4_c;
+    dsp->vc1_v_loop_filter8   = vc1_v_loop_filter8_c;
+    dsp->vc1_h_loop_filter8   = vc1_h_loop_filter8_c;
+    dsp->vc1_v_loop_filter16  = vc1_v_loop_filter16_c;
+    dsp->vc1_h_loop_filter16  = vc1_h_loop_filter16_c;
+
+    dsp->put_vc1_mspel_pixels_tab[0]  = put_vc1_mspel_mc00_c;
+    dsp->put_vc1_mspel_pixels_tab[1]  = put_vc1_mspel_mc10_c;
+    dsp->put_vc1_mspel_pixels_tab[2]  = put_vc1_mspel_mc20_c;
+    dsp->put_vc1_mspel_pixels_tab[3]  = put_vc1_mspel_mc30_c;
+    dsp->put_vc1_mspel_pixels_tab[4]  = put_vc1_mspel_mc01_c;
+    dsp->put_vc1_mspel_pixels_tab[5]  = put_vc1_mspel_mc11_c;
+    dsp->put_vc1_mspel_pixels_tab[6]  = put_vc1_mspel_mc21_c;
+    dsp->put_vc1_mspel_pixels_tab[7]  = put_vc1_mspel_mc31_c;
+    dsp->put_vc1_mspel_pixels_tab[8]  = put_vc1_mspel_mc02_c;
+    dsp->put_vc1_mspel_pixels_tab[9]  = put_vc1_mspel_mc12_c;
     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
@@ -643,20 +919,44 @@ av_cold void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
 
-    dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c;
-    dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c;
+    dsp->avg_vc1_mspel_pixels_tab[0]  = avg_vc1_mspel_mc00_c;
+    dsp->avg_vc1_mspel_pixels_tab[1]  = avg_vc1_mspel_mc10_c;
+    dsp->avg_vc1_mspel_pixels_tab[2]  = avg_vc1_mspel_mc20_c;
+    dsp->avg_vc1_mspel_pixels_tab[3]  = avg_vc1_mspel_mc30_c;
+    dsp->avg_vc1_mspel_pixels_tab[4]  = avg_vc1_mspel_mc01_c;
+    dsp->avg_vc1_mspel_pixels_tab[5]  = avg_vc1_mspel_mc11_c;
+    dsp->avg_vc1_mspel_pixels_tab[6]  = avg_vc1_mspel_mc21_c;
+    dsp->avg_vc1_mspel_pixels_tab[7]  = avg_vc1_mspel_mc31_c;
+    dsp->avg_vc1_mspel_pixels_tab[8]  = avg_vc1_mspel_mc02_c;
+    dsp->avg_vc1_mspel_pixels_tab[9]  = avg_vc1_mspel_mc12_c;
     dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;
     dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;
     dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;
     dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;
     dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;
     dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;
+
+    dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_c;
+    dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_c;
+    dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = put_no_rnd_vc1_chroma_mc4_c;
+    dsp->avg_no_rnd_vc1_chroma_pixels_tab[1] = avg_no_rnd_vc1_chroma_mc4_c;
+
+#if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER
+    dsp->sprite_h                 = sprite_h_c;
+    dsp->sprite_v_single          = sprite_v_single_c;
+    dsp->sprite_v_double_noscale  = sprite_v_double_noscale_c;
+    dsp->sprite_v_double_onescale = sprite_v_double_onescale_c;
+    dsp->sprite_v_double_twoscale = sprite_v_double_twoscale_c;
+#endif /* CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER */
+
+    dsp->startcode_find_candidate = ff_startcode_find_candidate_c;
+
+    if (ARCH_AARCH64)
+        ff_vc1dsp_init_aarch64(dsp);
+    if (ARCH_ARM)
+        ff_vc1dsp_init_arm(dsp);
+    if (ARCH_PPC)
+        ff_vc1dsp_init_ppc(dsp);
+    if (ARCH_X86)
+        ff_vc1dsp_init_x86(dsp);
 }