📄 mvmc.c
字号:
C = 16-RND - (Src[6]+Src[13]) + 3*(Src[ 7]+Src[12])-6*(Src[ 8]+Src[11]) + 20*(Src[ 9]+Src[10]);
CLIP_STORE( 9,C);
C = 16-RND - (Src[7]+Src[14]) + 3*(Src[ 8]+Src[13])-6*(Src[ 9]+Src[12]) + 20*(Src[10]+Src[11]);
CLIP_STORE(10,C);
C = 16-RND - (Src[8]+Src[15]) + 3*(Src[ 9]+Src[14])-6*(Src[10]+Src[13]) + 20*(Src[11]+Src[12]);
CLIP_STORE(11,C);
C = 16-RND - (Src[9]+Src[16]) + 3*(Src[10]+Src[15])-6*(Src[11]+Src[14]) + 20*(Src[12]+Src[13]);
CLIP_STORE(12,C);
C = 16-RND - Src[10] +3*Src[11] -6*(Src[12]+Src[15]) + 20*(Src[13]+Src[14]) +2*Src[16];
CLIP_STORE(13,C);
C = 16-RND - Src[11] +3*(Src[12]-Src[16]) -6*Src[13] + 20*Src[14] + 19*Src[15];
CLIP_STORE(14,C);
C = 16-RND - Src[12] +3*Src[13] -7*Src[14] + 23*Src[15] + 14*Src[16];
CLIP_STORE(15,C);
Src += BpS;
Dst += BpS;
}
}
#undef CLIP_STORE
#define CLIP_STORE(i,C) \
if (C<0) C = 0; else if (C>(255<<5)) C = 255; else C = C>>5; \
C = (C+Src[i+1]+1-RND) >> 1; \
STORE(Dst[i], C)
void interpolate16x16_quarterpel_ha_up_add(uint8_t *Dst, const uint8_t *Src, int32_t H, int32_t BpS, int32_t RND)
{
while(H-->0)
{
int C;
C = 16-RND +14*Src[0] +23*Src[1] - 7*Src[2] + 3*Src[3] - Src[4];
CLIP_STORE(0,C);
C = 16-RND - 3*(Src[0]-Src[4]) +19*Src[1] +20*Src[2] - 6*Src[3] - Src[5];
CLIP_STORE( 1,C);
C = 16-RND + 2*Src[0] - 6*(Src[1]+Src[4]) +20*(Src[2]+Src[3]) + 3*Src[5] - Src[6];
CLIP_STORE( 2,C);
C = 16-RND - (Src[0]+Src[7 ]) + 3*(Src[ 1]+Src[ 6])-6*(Src[ 2]+Src[ 5]) + 20*(Src[ 3]+Src[ 4]);
CLIP_STORE( 3,C);
C = 16-RND - (Src[1]+Src[8 ]) + 3*(Src[ 2]+Src[ 7])-6*(Src[ 3]+Src[ 6]) + 20*(Src[ 4]+Src[ 5]);
CLIP_STORE( 4,C);
C = 16-RND - (Src[2]+Src[9 ]) + 3*(Src[ 3]+Src[ 8])-6*(Src[ 4]+Src[ 7]) + 20*(Src[ 5]+Src[ 6]);
CLIP_STORE( 5,C);
C = 16-RND - (Src[3]+Src[10]) + 3*(Src[ 4]+Src[ 9])-6*(Src[ 5]+Src[ 8]) + 20*(Src[ 6]+Src[ 7]);
CLIP_STORE( 6,C);
C = 16-RND - (Src[4]+Src[11]) + 3*(Src[ 5]+Src[10])-6*(Src[ 6]+Src[ 9]) + 20*(Src[ 7]+Src[ 8]);
CLIP_STORE( 7,C);
C = 16-RND - (Src[5]+Src[12]) + 3*(Src[ 6]+Src[11])-6*(Src[ 7]+Src[10]) + 20*(Src[ 8]+Src[ 9]);
CLIP_STORE( 8,C);
C = 16-RND - (Src[6]+Src[13]) + 3*(Src[ 7]+Src[12])-6*(Src[ 8]+Src[11]) + 20*(Src[ 9]+Src[10]);
CLIP_STORE( 9,C);
C = 16-RND - (Src[7]+Src[14]) + 3*(Src[ 8]+Src[13])-6*(Src[ 9]+Src[12]) + 20*(Src[10]+Src[11]);
CLIP_STORE(10,C);
C = 16-RND - (Src[8]+Src[15]) + 3*(Src[ 9]+Src[14])-6*(Src[10]+Src[13]) + 20*(Src[11]+Src[12]);
CLIP_STORE(11,C);
C = 16-RND - (Src[9]+Src[16]) + 3*(Src[10]+Src[15])-6*(Src[11]+Src[14]) + 20*(Src[12]+Src[13]);
CLIP_STORE(12,C);
C = 16-RND - Src[10] +3*Src[11] -6*(Src[12]+Src[15]) + 20*(Src[13]+Src[14]) +2*Src[16];
CLIP_STORE(13,C);
C = 16-RND - Src[11] +3*(Src[12]-Src[16]) -6*Src[13] + 20*Src[14] + 19*Src[15];
CLIP_STORE(14,C);
C = 16-RND - Src[12] +3*Src[13] -7*Src[14] + 23*Src[15] + 14*Src[16];
CLIP_STORE(15,C);
Src += BpS;
Dst += BpS;
}
}
#undef CLIP_STORE
#define CLIP_STORE(D,C) \
if (C<0) C = 0; else if (C>(255<<5)) C = 255; else C = C>>5; \
STORE(D, C)
void interpolate16x16_quarterpel_v_add(uint8_t *Dst, const uint8_t *Src, int32_t H, int32_t BpS, int32_t RND)
{
int C, p, q, s, t, w, m, n, l;
while(H-->0)
{
p = 0; q = p + BpS; s = q + BpS; t = s + BpS; w = t + BpS;
C = 16-RND +14*Src[p] +23*Src[q] - 7*Src[s] + 3*Src[t] - Src[w];
CLIP_STORE(Dst[p],C);
m = w + BpS;
C = 16-RND - 3*(Src[p]-Src[w]) +19*Src[q] +20*Src[s] - 6*Src[t] - Src[m];
CLIP_STORE(Dst[q],C);
n = m + BpS;
C = 16-RND + 2*Src[p] - 6*(Src[q]+Src[w]) +20*(Src[s]+Src[t]) + 3*Src[m] - Src[n];
CLIP_STORE(Dst[s],C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(Dst[t],C);
p = l + BpS;
C = 16-RND - (Src[q]+Src[p ]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(Dst[w],C);
q = p + BpS;
C = 16-RND - (Src[s]+Src[q ]) + 3*(Src[t]+Src[p])-6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]);
CLIP_STORE(Dst[m],C);
s = q + BpS;
C = 16-RND - (Src[t]+Src[s]) + 3*(Src[w]+Src[q])-6*(Src[m]+Src[p]) + 20*(Src[n]+Src[l]);
CLIP_STORE(Dst[n],C);
t = s + BpS;
C = 16-RND - (Src[w]+Src[t]) + 3*(Src[m]+Src[s])-6*(Src[n]+Src[q]) + 20*(Src[l]+Src[p]);
CLIP_STORE(Dst[l],C);
w = t + BpS;
C = 16-RND - (Src[m]+Src[w]) + 3*(Src[n]+Src[t])-6*(Src[l]+Src[s]) + 20*(Src[p]+Src[q]);
CLIP_STORE(Dst[p],C);
m = w + BpS;
C = 16-RND - (Src[n]+Src[m]) + 3*(Src[l]+Src[w])-6*(Src[p]+Src[t]) + 20*(Src[q]+Src[s]);
CLIP_STORE(Dst[q],C);
n = m + BpS;
C = 16-RND - (Src[l]+Src[n]) + 3*(Src[p]+Src[m])-6*(Src[q]+Src[w]) + 20*(Src[s]+Src[t]);
CLIP_STORE(Dst[s],C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(Dst[t],C);
p = l +BpS;
C = 16-RND - (Src[q]+Src[p]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(Dst[w],C);
C = 16-RND - Src[s] +3*Src[t] -6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]) +2*Src[p];
CLIP_STORE(Dst[m],C);
C = 16-RND - Src[t] +3*(Src[w]-Src[p]) -6*Src[m] + 20*Src[n] + 19*Src[l];
CLIP_STORE(Dst[n],C);
C = 16-RND - Src[w] +3*Src[m] -7*Src[n] + 23*Src[l] + 14*Src[p];
CLIP_STORE(Dst[l],C);
Src += 1;
Dst += 1;
}
}
#undef CLIP_STORE
#define CLIP_STORE(i,C) \
if (C<0) C = 0; else if (C>(255<<5)) C = 255; else C = C>>5; \
C = (C+Src[BpS*i]+1-RND) >> 1; \
STORE(Dst[BpS*i], C)
void interpolate16x16_quarterpel_va_add(uint8_t *Dst, const uint8_t *Src, int32_t H, int32_t BpS, int32_t RND)
{
int C, p, q, s, t, w, m, n, l;
while(H-->0)
{
p = 0; q = p + BpS; s = q + BpS; t = s + BpS; w = t + BpS;
C = 16-RND +14*Src[p] +23*Src[q] - 7*Src[s] + 3*Src[t] - Src[w];
CLIP_STORE(0,C);
m = w + BpS;
C = 16-RND - 3*(Src[p]-Src[w]) +19*Src[q] +20*Src[s] - 6*Src[t] - Src[m];
CLIP_STORE(1,C);
n = m + BpS;
C = 16-RND + 2*Src[p] - 6*(Src[q]+Src[w]) +20*(Src[s]+Src[t]) + 3*Src[m] - Src[n];
CLIP_STORE(2,C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(3,C);
p = l + BpS;
C = 16-RND - (Src[q]+Src[p ]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(4,C);
q = p + BpS;
C = 16-RND - (Src[s]+Src[q ]) + 3*(Src[t]+Src[p])-6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]);
CLIP_STORE(5,C);
s = q + BpS;
C = 16-RND - (Src[t]+Src[s]) + 3*(Src[w]+Src[q])-6*(Src[m]+Src[p]) + 20*(Src[n]+Src[l]);
CLIP_STORE(6,C);
t = s + BpS;
C = 16-RND - (Src[w]+Src[t]) + 3*(Src[m]+Src[s])-6*(Src[n]+Src[q]) + 20*(Src[l]+Src[p]);
CLIP_STORE(7,C);
w = t + BpS;
C = 16-RND - (Src[m]+Src[w]) + 3*(Src[n]+Src[t])-6*(Src[l]+Src[s]) + 20*(Src[p]+Src[q]);
CLIP_STORE(8,C);
m = w + BpS;
C = 16-RND - (Src[n]+Src[m]) + 3*(Src[l]+Src[w])-6*(Src[p]+Src[t]) + 20*(Src[q]+Src[s]);
CLIP_STORE(9,C);
n = m + BpS;
C = 16-RND - (Src[l]+Src[n]) + 3*(Src[p]+Src[m])-6*(Src[q]+Src[w]) + 20*(Src[s]+Src[t]);
CLIP_STORE(10,C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(11,C);
p = l +BpS;
C = 16-RND - (Src[q]+Src[p]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(12,C);
C = 16-RND - Src[s] +3*Src[t] -6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]) +2*Src[p];
CLIP_STORE(13,C);
C = 16-RND - Src[t] +3*(Src[w]-Src[p]) -6*Src[m] + 20*Src[n] + 19*Src[l];
CLIP_STORE(14,C);
C = 16-RND - Src[w] +3*Src[m] -7*Src[n] + 23*Src[l] + 14*Src[p];
CLIP_STORE(15,C);
Src += 1;
Dst += 1;
}
}
#undef CLIP_STORE
#define CLIP_STORE(i,C) \
if (C<0) C = 0; else if (C>(255<<5)) C = 255; else C = C>>5; \
C = (C+Src[BpS*i+BpS]+1-RND) >> 1; \
STORE(Dst[BpS*i], C)
void interpolate16x16_quarterpel_va_up_add(uint8_t *Dst, const uint8_t *Src, int32_t H, int32_t BpS, int32_t RND)
{
int C, p, q, s, t, w, m, n, l;
while(H-->0)
{
p = 0; q = p + BpS; s = q + BpS; t = s + BpS; w = t + BpS;
C = 16-RND +14*Src[p] +23*Src[q] - 7*Src[s] + 3*Src[t] - Src[w];
CLIP_STORE(0,C);
m = w + BpS;
C = 16-RND - 3*(Src[p]-Src[w]) +19*Src[q] +20*Src[s] - 6*Src[t] - Src[m];
CLIP_STORE(1,C);
n = m + BpS;
C = 16-RND + 2*Src[p] - 6*(Src[q]+Src[w]) +20*(Src[s]+Src[t]) + 3*Src[m] - Src[n];
CLIP_STORE(2,C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(3,C);
p = l + BpS;
C = 16-RND - (Src[q]+Src[p ]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(4,C);
q = p + BpS;
C = 16-RND - (Src[s]+Src[q ]) + 3*(Src[t]+Src[p])-6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]);
CLIP_STORE(5,C);
s = q + BpS;
C = 16-RND - (Src[t]+Src[s]) + 3*(Src[w]+Src[q])-6*(Src[m]+Src[p]) + 20*(Src[n]+Src[l]);
CLIP_STORE(6,C);
t = s + BpS;
C = 16-RND - (Src[w]+Src[t]) + 3*(Src[m]+Src[s])-6*(Src[n]+Src[q]) + 20*(Src[l]+Src[p]);
CLIP_STORE(7,C);
w = t + BpS;
C = 16-RND - (Src[m]+Src[w]) + 3*(Src[n]+Src[t])-6*(Src[l]+Src[s]) + 20*(Src[p]+Src[q]);
CLIP_STORE(8,C);
m = w + BpS;
C = 16-RND - (Src[n]+Src[m]) + 3*(Src[l]+Src[w])-6*(Src[p]+Src[t]) + 20*(Src[q]+Src[s]);
CLIP_STORE(9,C);
n = m + BpS;
C = 16-RND - (Src[l]+Src[n]) + 3*(Src[p]+Src[m])-6*(Src[q]+Src[w]) + 20*(Src[s]+Src[t]);
CLIP_STORE(10,C);
l = n + BpS;
C = 16-RND - (Src[p]+Src[l]) + 3*(Src[q]+Src[n])-6*(Src[s]+Src[m]) + 20*(Src[t]+Src[w]);
CLIP_STORE(11,C);
p = l +BpS;
C = 16-RND - (Src[q]+Src[p]) + 3*(Src[s]+Src[l])-6*(Src[t]+Src[n]) + 20*(Src[w]+Src[m]);
CLIP_STORE(12,C);
C = 16-RND - Src[s] +3*Src[t] -6*(Src[w]+Src[l]) + 20*(Src[m]+Src[n]) +2*Src[p];
CLIP_STORE(13,C);
C = 16-RND - Src[t] +3*(Src[w]-Src[p]) -6*Src[m] + 20*Src[n] + 19*Src[l];
CLIP_STORE(14,C);
C = 16-RND - Src[w] +3*Src[m] -7*Src[n] + 23*Src[l] + 14*Src[p];
CLIP_STORE(15,C);
Src += 1;
Dst += 1;
}
}
#undef CLIP_STORE
void interpolate16x16_quarterpel_add(uint8_t * const cur,
uint8_t * const refn,
uint8_t * const refh,
uint8_t * const refv,
uint8_t * const refhv,
const uint32_t x, const uint32_t y,
const int32_t dx, const int dy,
const uint32_t stride,
const uint32_t rounding)
{
const uint8_t *src;
uint8_t *dst;
uint8_t *tmp;
int32_t quads;
int32_t x_int, y_int;
const int32_t xRef = (int)x*4 + dx;
const int32_t yRef = (int)y*4 + dy;
quads = (dx&3) | ((dy&3)<<2);
x_int = xRef >> 2;
y_int = yRef >> 2;
dst = cur + y * stride + x;
src = refn + y_int * (int)stride + x_int;
tmp = refh; /* we need at least a 16 x stride scratch block */
switch(quads)
{
case 0:
interpolate8x8_halfpel_add(dst, dst, src, stride, 0, 8);
interpolate8x8_halfpel_add(dst+8, dst+8, src+8, stride, 0, 8);
interpolate8x8_halfpel_add(dst+8*stride, dst+8*stride, src+8*stride, stride, 0, 8);
interpolate8x8_halfpel_add(dst+8*stride+8, dst+8*stride+8, src+8*stride+8, stride, 0, 8);
break;
case 1:
interpolate16x16_quarterpel_ha_add(dst, src, 16, stride, rounding);
break;
case 2:
interpolate16x16_quarterpel_h_add(dst, src, 16, stride, rounding);
break;
case 3:
interpolate16x16_quarterpel_ha_up_add(dst, src, 16, stride, rounding);
break;
case 4:
interpolate16x16_quarterpel_va_add(dst, src, 16, stride, rounding);
break;
case 5:
interpolate16x16_quarterpel_ha(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_va_add(dst, tmp, 16, stride, rounding);
break;
case 6:
interpolate16x16_quarterpel_h(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_va_add(dst, tmp, 16, stride, rounding);
break;
case 7:
interpolate16x16_quarterpel_ha_up(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_va_add(dst, tmp, 16, stride, rounding);
break;
case 8:
interpolate16x16_quarterpel_v_add(dst, src, 16, stride, rounding);
break;
case 9:
interpolate16x16_quarterpel_ha(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_v_add(dst, tmp, 16, stride, rounding);
break;
case 10:
interpolate16x16_quarterpel_h(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_v_add(dst, tmp, 16, stride, rounding);
break;
case 11:
interpolate16x16_quarterpel_ha_up(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_v_add(dst, tmp, 16, stride, rounding);
break;
case 12:
interpolate16x16_quarterpel_va_up_add(dst, src, 16, stride, rounding);
break;
case 13:
interpolate16x16_quarterpel_ha(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_va_up_add(dst, tmp, 16, stride, rounding);
break;
case 14:
interpolate16x16_quarterpel_h(tmp, src, 17, stride, rounding);
interpolate16x16_quarterpel_va_up_add( dst, tmp, 16, stride, rounding);
break;
case 15:
interpolat⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -