CANCELLED: POSTER: Evaluating AMD GPUs for Overlap-Save Convolutions in Time Domain Radio Astronomy

GPU implementations for core astronomical libraries

In this poster, we will present a case study of converting our GPU accelerated convolution code using the overlap-save algorithm to use AMD GPUs. We will cover the experience of converting the code from CUDA to AMD's HIP, and compare the performance of the HIP version to the CUDA version on both NVIDIA and AMD GPUs.