Copyrights to these papers may be held by the publishers. The download files are preprints. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
Anuva Kulkarni, Franz Franchetti and Jelena Kovacevic (, 2018)
Algorithm Design for Large Scale FFT-Based Simulations on CPU-GPU Platforms
Comment: poster with extended abstract
Preprint (1.2 MB)
Published paper (link to publisher)
Extreme memory requirements and high communication overhead prevent scaling of large scale iterative simulations involving parallel Fast Fourier Transforms (FFTs) to higher grid sizes, which is necessary for high resolution analysis. To overcome these limitations, we propose an algorithm to run stress-strain simulations on CPU-GPU platforms for larger problem sizes using irregular domain decomposition and local FFTs. Early results show that our method lowers iteration cost without adversely impacting accuracy of the result.Keywords: Fast Fourier Transform, CPUs, Simulation, Design, GPUs, Algorithm, Large Scale