Minimum conflict algorithm was very slow for me. But I could optimize it using intensive parallelism and an observation to count straight line conflicts.

The observation is, if and are two lattice points, then an other lattice point on the straight line would have this form,

Using this would reduce number of iteration in huge factor. Parallelism helps because most of the work in each iteration is independent.

Using this, I was able to find a solution within one hour.