void transform_skip_8_fallback(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
{
int nT = 4;
int bdShift2 = 20-8;
for (int y=0;y<nT;y++)
for (int x=0;x<nT;x++) {
int32_t c = coeffs[x+y*nT] << 7;
c = (c+(1<<(bdShift2-1)))>>bdShift2;
dst[y*stride+x] = Clip1_8bit(dst[y*stride+x] + c);
}
}
void transform_4x4_luma_add_8_fallback(uint8_t *dst, int16_t *coeffs, ptrdiff_t stride)
{
int16_t g[4][4];
int postShift = 20-8; // 8 bit
int rndV = 1<<(7-1);
int rndH = 1<<(postShift-1);
// --- V ---
for (int c=0;c<4;c++) {
logtrace(LogTransform,"DST-V: ");
for (int r=0;r<4;r++) {
logtrace(LogTransform,"%d ",coeffs[c+r*4]);
}
logtrace(LogTransform,"* -> ");
for (int i=0;i<4;i++) {
int sum=0;
for (int j=0;j<4;j++) {
sum += mat_8_357[j][i] * coeffs[c+j*4];
}
g[i][c] = Clip3(-32768,32767, (sum+rndV)>>7);
}
for (int y=0;y<4;y++) {
logtrace(LogTransform,"*%d ",g[y][c]);
}
logtrace(LogTransform,"*\n");
}
// --- H ---
for (int y=0;y<4;y++) {
logtrace(LogTransform,"DST-H: ");
for (int c=0;c<4;c++) {
logtrace(LogTransform,"%d ",g[y][c]);
}
logtrace(LogTransform,"* -> ");
for (int i=0;i<4;i++) {
int sum=0;
for (int j=0;j<4;j++) {
sum += mat_8_357[j][i] * g[y][j];
}
int out = Clip3(-32768,32767, (sum+rndH)>>postShift);
dst[y*stride+i] = Clip1_8bit(dst[y*stride+i] + out);
logtrace(LogTransform,"*%d ",out);
}
logtrace(LogTransform,"*\n");
}
}
static void transform_dct_add_8(uint8_t *dst, ptrdiff_t stride,
int nT, int16_t *coeffs)
{
int postShift = 20-8; // 8 bit
int rnd1 = 1<<(7-1);
int rnd2 = 1<<(postShift-1);
int fact = (1<<(5-Log2(nT)));
int16_t g[32*32]; // actually, only [nT*nT] used
// TODO: valgrind reports that dst[] contains uninitialized data.
// Probably from intra-prediction.
/*
for (int i=0;i<nT*nT;i++) {
printf("%d\n",coeffs[i]);
}
for (int y=0;y<nT;y++) {
for (int i=0;i<nT;i++) {
printf("%d ",dst[y*stride+i]);
}
}
printf("\n");
*/
for (int c=0;c<nT;c++) {
logtrace(LogTransform,"DCT-V: ");
for (int i=0;i<nT;i++) {
logtrace(LogTransform,"*%d ",coeffs[c+i*nT]);
}
logtrace(LogTransform,"* -> ");
// find last non-zero coefficient to reduce computations carried out in DCT
int lastCol = nT-1;
for (;lastCol>=0;lastCol--) {
if (coeffs[c+lastCol*nT]) { break; }
}
for (int i=0;i<nT;i++) {
int sum=0;
for (int j=0;j<=lastCol /*nT*/;j++) {
sum += mat_dct[fact*j][i] * coeffs[c+j*nT];
}
g[c+i*nT] = Clip3(-32768,32767, (sum+rnd1)>>7);
logtrace(LogTransform,"*%d ",g[c+i*nT]);
}
logtrace(LogTransform,"*\n");
}
for (int y=0;y<nT;y++) {
logtrace(LogTransform,"DCT-H: ");
for (int i=0;i<nT;i++) {
logtrace(LogTransform,"*%d ",g[i+y*nT]);
}
logtrace(LogTransform,"* -> ");
// find last non-zero coefficient to reduce computations carried out in DCT
int lastCol = nT-1;
for (;lastCol>=0;lastCol--) {
if (g[y*nT+lastCol]) { break; }
}
for (int i=0;i<nT;i++) {
int sum=0;
for (int j=0;j<=lastCol /*nT*/;j++) {
sum += mat_dct[fact*j][i] * g[y*nT+j];
}
//int out = Clip3(-32768,32767, (sum+rnd2)>>postShift);
int out = (sum+rnd2)>>postShift;
//printf("%d\n",Clip1_8bit(dst[y*stride+i]));
//printf("%d\n",Clip1_8bit(out));
dst[y*stride+i] = Clip1_8bit(dst[y*stride+i] + out);
logtrace(LogTransform,"*%d ",out);
}
logtrace(LogTransform,"*\n");
}
}
static void transform_dct_add_8(uint8_t *dst, ptrdiff_t stride,
int nT, int16_t *coeffs)
{
int postShift = 20-8; // 8 bit
int rnd1 = 1<<(7-1);
int rnd2 = 1<<(postShift-1);
int fact = (1<<(5-Log2(nT)));
int16_t g[32*32]; // actually, only [nT*nT] used
//int16_t col[32]; // actually, only [nT] used
//int32_t out[32]; // actually, only [nT] used
for (int c=0;c<nT;c++) {
logtrace(LogTransform,"DCT-V: ");
for (int i=0;i<nT;i++) {
logtrace(LogTransform,"*%d ",coeffs[c+i*nT]);
}
logtrace(LogTransform,"* -> ");
// find last non-zero coefficient to reduce computations carried out in DCT
int lastCol = nT-1;
for (;lastCol>=0;lastCol--) {
if (coeffs[c+lastCol*nT]) { break; }
}
for (int i=0;i<nT;i++) {
int sum=0;
for (int j=0;j<=lastCol /*nT*/;j++) {
sum += mat_dct[fact*j][i] * coeffs[c+j*nT];
}
g[c+i*nT] = Clip3(-32768,32767, (sum+rnd1)>>7);
logtrace(LogTransform,"*%d ",g[c+i*nT]);
}
logtrace(LogTransform,"*\n");
}
for (int y=0;y<nT;y++) {
logtrace(LogTransform,"DCT-H: ");
for (int i=0;i<nT;i++) {
logtrace(LogTransform,"*%d ",g[i+y*nT]);
}
logtrace(LogTransform,"* -> ");
// find last non-zero coefficient to reduce computations carried out in DCT
int lastCol = nT-1;
for (;lastCol>=0;lastCol--) {
if (g[y*nT+lastCol]) { break; }
}
for (int i=0;i<nT;i++) {
int sum=0;
for (int j=0;j<=lastCol /*nT*/;j++) {
sum += mat_dct[fact*j][i] * g[y*nT+j];
}
//int out = Clip3(-32768,32767, (sum+rnd2)>>postShift);
int out = (sum+rnd2)>>postShift;
dst[y*stride+i] = Clip1_8bit(dst[y*stride+i] + out);
logtrace(LogTransform,"*%d ",out);
}
logtrace(LogTransform,"*\n");
}
}
