Roee Zisholz

Fix exploding values

1

Fix non convergence of the local smoothing filter iterations, as sdot(d,q) would be Infinity for large numbers, eventually resulting in NaN rnorm. the fix changes float to 64bit double precision...