void get_inter_prediction_luma(uint8_t *pblock, uint8_t *ref, int width, int height, int stride, int pstride, mv_t *mv, int sign, int bipred)
{
int i,j;
int m,i_off,j_off;
mv_t mvtemp;
mvtemp.x = sign ? -mv->x : mv->x;
mvtemp.y = sign ? -mv->y : mv->y;
int ver_frac = (mvtemp.y)&3;
int hor_frac = (mvtemp.x)&3;
int ver_int = (mvtemp.y)>>2;
int hor_int = (mvtemp.x)>>2;
int32_t tmp[MAX_BLOCK_SIZE+16][MAX_BLOCK_SIZE + 16]; //7-bit filter exceeds 16 bit temporary storage
/* Integer position */
if (ver_frac==0 && hor_frac==0){
j_off = 0 + hor_int;
for(i=0;i<height;i++){
i_off = i + ver_int;
memcpy(pblock + i*pstride,ref + i_off*stride+j_off, width*sizeof(uint8_t));
}
return;
}
if (use_simd)
get_inter_prediction_luma_simd(width, height, hor_frac, ver_frac, pblock, pstride, ref + ver_int*stride + hor_int, stride, bipred);
/* Special lowpass filter at center position */
else if (ver_frac == 2 && hor_frac == 2) {
for(i=0;i<height;i++){
for (j=0;j<width;j++){
int sum = 0;
i_off = i + ver_int;
j_off = j + hor_int;
sum += 0*ref[(i_off-1)*stride+j_off-1]+1*ref[(i_off-1)*stride+j_off+0]+1*ref[(i_off-1)*stride+j_off+1]+0*ref[(i_off-1)*stride+j_off+2];
sum += 1*ref[(i_off+0)*stride+j_off-1]+2*ref[(i_off+0)*stride+j_off+0]+2*ref[(i_off+0)*stride+j_off+1]+1*ref[(i_off+0)*stride+j_off+2];
sum += 1*ref[(i_off+1)*stride+j_off-1]+2*ref[(i_off+1)*stride+j_off+0]+2*ref[(i_off+1)*stride+j_off+1]+1*ref[(i_off+1)*stride+j_off+2];
sum += 0*ref[(i_off+2)*stride+j_off-1]+1*ref[(i_off+2)*stride+j_off+0]+1*ref[(i_off+2)*stride+j_off+1]+0*ref[(i_off+2)*stride+j_off+2];
pblock[i*pstride+j] = clip255((sum + 8)>>4);
}
}
} else {
void get_inter_prediction_luma(uint8_t *pblock, uint8_t *ref, int width, int height, int stride, int pstride, mv_t *mv, int sign)
{
int i,j;
int m,i_off,j_off;
mv_t mvtemp;
mvtemp.x = sign ? -mv->x : mv->x;
mvtemp.y = sign ? -mv->y : mv->y;
int ver_frac = (mvtemp.y)&3;
int hor_frac = (mvtemp.x)&3;
int ver_int = (mvtemp.y)>>2;
int hor_int = (mvtemp.x)>>2;
int32_t tmp[80][80]; //7-bit filter exceeds 16 bit temporary storage
/* Integer position */
if (ver_frac==0 && hor_frac==0){
j_off = 0 + hor_int;
for(i=0;i<height;i++){
i_off = i + ver_int;
memcpy(pblock + i*pstride,ref + i_off*stride+j_off, width*sizeof(uint8_t));
}
return;
}
#if HEVC_INTERPOLATION
/* Vertical filtering */
for(i=-OFFYM1;i<width+OFFY;i++){
for (j=0;j<height;j++){
int sum = 0;
i_off = i + hor_int;
j_off = j + ver_int;
for (m=0;m<NTAPY;m++) sum += filter_coeffsY[ver_frac][m] * ref[(j_off + m - OFFYM1) * stride + i_off];
tmp[j][i+OFFYM1] = sum;
}
}
/* Horizontal filtering */
for(i=0;i<width;i++){
for (j=0;j<height;j++){
int sum = 0;
for (m=0;m<NTAPY;m++) sum += filter_coeffsY[hor_frac][m] * tmp[j][i+m];
pblock[j*pstride+i] = clip255((sum + 2048)>>12);
}
}
return;
#endif
if (use_simd) {
get_inter_prediction_luma_simd(width, height, hor_frac, ver_frac, pblock, pstride, ref + ver_int*stride + hor_int, stride);
}
else {
/* Special lowpass filter at center position */
if (ver_frac == 2 && hor_frac == 2) {
for(i=0;i<height;i++){
for (j=0;j<width;j++){
int sum = 0;
i_off = i + ver_int;
j_off = j + hor_int;
sum += 0*ref[(i_off-1)*stride+j_off-1]+1*ref[(i_off-1)*stride+j_off+0]+1*ref[(i_off-1)*stride+j_off+1]+0*ref[(i_off-1)*stride+j_off+2];
sum += 1*ref[(i_off+0)*stride+j_off-1]+2*ref[(i_off+0)*stride+j_off+0]+2*ref[(i_off+0)*stride+j_off+1]+1*ref[(i_off+0)*stride+j_off+2];
sum += 1*ref[(i_off+1)*stride+j_off-1]+2*ref[(i_off+1)*stride+j_off+0]+2*ref[(i_off+1)*stride+j_off+1]+1*ref[(i_off+1)*stride+j_off+2];
sum += 0*ref[(i_off+2)*stride+j_off-1]+1*ref[(i_off+2)*stride+j_off+0]+1*ref[(i_off+2)*stride+j_off+1]+0*ref[(i_off+2)*stride+j_off+2];
pblock[i*pstride+j] = clip255((sum + 8)>>4);
}
}
} else {
/* Vertical filtering */
for(i=-OFFYM1;i<width+OFFY;i++){
for (j=0;j<height;j++){
int sum = 0;
i_off = i + hor_int;
j_off = j + ver_int;
for (m=0;m<NTAPY;m++) sum += filter_coeffsY[ver_frac][m] * ref[(j_off + m - OFFYM1) * stride + i_off]; //7-bit version
tmp[j][i+OFFYM1] = sum;
}
}
/* Horizontal filtering */
for(i=0;i<width;i++){
for (j=0;j<height;j++){
int sum = 0;
for (m=0;m<NTAPY;m++) sum += filter_coeffsY[hor_frac][m] * tmp[j][i+m]; //7-bit version
pblock[j*pstride+i] = clip255((sum + 8192)>>14); //7-bit version
}
}
}
}
