Yuri Dotsenko, Sara S. Baghsorkhi, Brandon Lloyd, and Naga K. Govindaraju
We present an auto-tuning framework for FFTs on graphics processors (GPUs). Due to complex design of the memory and compute subsystems on GPUs, the performance of FFT kernels over the range of possible input parameters can vary widely. We generate several variants for each component of the FFT kernel that, for different cases, are likely to perform well. Our auto-tuner composes variants to generate kernels and selects the best ones. We present heuristics to prune the search space and profile only a small fraction of all possible kernels. We compose optimized kernels to improve the performance of larger FFT computations. We implement the system using the NVIDIA CUDA API and compare its performance to the state-of-the-art FFT libraries. On a range of NVIDIA GPUs and input sizes, our auto-tuned FFTs outperform the NVIDIA CUFFT 3.0 library by up to 38x and deliver up to 3x higher performance compared to a manually-tuned FFT.
In Proceedings of the 16th ACM symposium on Principles and practice of parallel programming