Open
Description
Determining the magnetic field takes a lot of time because it is CPU-only. As such, a GPU kernel would be nice... Along with an additional derivative function (psuedocode):
___global___ void derive(data, out, stride, n, dx){
int gid = [init]
if (gid + stride < n){
out[gid] = (data[gid+stride] - data[gid])/dx
}
else{
out[gid] = data[gid]/dx
}
}
Here, a stride of 1 -> xDim, xDim -> yDim, xDim*yDim -> zDim derivatives
then we just use this to create our curl function, so we might want to make a separate __device__
kernel as well.
Metadata
Assignees
Labels
No labels