Embedded Files
High performance code for Nvidia GPUs
2D & 3D Heat Diffusion Modeling
This CUDA code solves the heat equation for a defined area or volume utilizing a dynamic number of GPU threads for maximum efficiency. All parallelizable code is executed on the GPU as opposed to the CPU increasing efficiency by nearly 500%.
The modeling equations used for computation
2D Discrete Fourier Transform
This CUDA code computes the 2D DFT with a dynamic number of threads for optimal efficiency. For this particular implementation, the original formula was used
Despite not implementing the improved Fast Fourier Transform algorithm, this CUDA code still out-performed an FFT implementation using MPI. Compare the run times of some different implementations below
Page updated
Google Sites
Report abuse