IRIS Digital Asset: Random number generation for high-energy physics simulation with FPGAs
Lead Research Organisation:
Imperial College London
Department Name: Physics
Abstract
This grant has been awarded to complete a set scope of work for IRIS.
First we will investigate, in collaboration with the original authors of mixmax, how it can be ported to an FPGA. The procedure will be to gain a deep understanding of how FPGA design constraints, features and optimisations can be integrated in the original algorithm and produce a computational model as an essential step toward FPGA implementation.
Second, we will produce an FPGA implementation of the adapted algorithm and benchmark its performance in terms of quality of the random numbers, latency and resource usage in contemporary devices. The deliverable will be an open-source code implementation of the algorithm and potentially a conference presentation or paper.
First we will investigate, in collaboration with the original authors of mixmax, how it can be ported to an FPGA. The procedure will be to gain a deep understanding of how FPGA design constraints, features and optimisations can be integrated in the original algorithm and produce a computational model as an essential step toward FPGA implementation.
Second, we will produce an FPGA implementation of the adapted algorithm and benchmark its performance in terms of quality of the random numbers, latency and resource usage in contemporary devices. The deliverable will be an open-source code implementation of the algorithm and potentially a conference presentation or paper.