Hi , is anyone familiar with EM simulation on CUDA?
A 1-D FDTD was implemented in CUDA, and runs well.
But when I try to ues share memory, some easy boundary
coundition E(0)=0, E(Nz)=0. seems to lose effect of reflecting wave back.
(main kernel below)
int idx = blockIdx.x*blockDim.x+threadIdx.x;
if ( idx < Nz) Eout[idx+1] = s_E[idx+1]- (s_H[idx+1]-s_H[idx]);
Eout = 0;