Anyone implemented the Gerchberg-Saxton Algorithm on a GPU?

[Ben_Hughes]

I am completely new to GPU programming. I want to implement the Gerchberg-Saxton algorithm on a GPU. I have a version running in LabVIEW on teh CPU, but it takes about 0.3 seconds per itteration, which is too slow. The algorithm uses FFT and InverseFFT and some manipulation of phase and amplitudes from complex data. Has anyone implemented this algorithm in CUDA and interfaced it to LabVIEW? If so, could you please give me some pointers?

Thanks

Ben

[biophysics_bob]

Are you familiar with C/C++ programming? If you can implement your algorithm in C or C++, then it is very simple to compile this program as a DLL, and interface it with Labview using the Call External Function node. We (mainly an undergraduate in our lab named Shawn) did this to write image processing algorithm, which uses some Fourier transforms as part of the particle tracking routine.

A working version of our program including a Labview demo and all the source code is available here:

http://www.engr.ucsb.edu/~saleh/