Hi,
Can anyone help me in coverting this kind of loop in Cuda with high performance…
consider abc and x both as floating point.
Length = 120,
iter = 8
abc[0][0] = 0;
for(int m = 1; m < iter; m++)
abc[0][m] = 0;
for(int k = 1; k <= Length; k++)
{
x[k] = 0;
for(m = 0; m < iter; m++)
{
abc[k][m] = (abc[k-1][m] + y[k-1][m])>( abc[k-1][m] + y[k-1][m]) ? (abc[k-1][m] + y[k-1][m]) : ( abc[k-1][m] + y[k-1][m]);
x[k] = x[k]> abc[k][m] ? x[k] : abc[k][m];
}
for(m = 0; m < iter; m++)
{
abc[k][m] = abc[k][m]-x[k];
}
}
Thanks…!