Guest
Presented by Abhay Samant and David Asplund at GNU Radio Conference 2020 GPUs are becoming increasingly popular as the compute platform for digital signal processing algorithms in cognitive radar and EW applications. Traditionally, it has been challenging to combine the high functionality UHD drivers running on a CPU with the high-performance algorithms running on a GPU. To do so, users would have to develop a user-level application to share data between the two processors. This would lead to multiple memory copies, non-determinism, increased CPU usage and non-optimal use of host controller memory bandwidth and space. In this presentation, we will present our findings on state-of-the-art processors from two major vendors and compare their performance for our application.
