Critique of “MemXCT: Memory-Centric X-Ray CT Reconstruction With Massive Parallelization” by SCC Team From Clemson University

Abstract

This paper reproduces the results of the Supercomputing 2019 paper, MemXCT: Memory-Centric X-ray CT Reconstruction with Massive Parallelization by Hidayeto˘ glu et al. as part of the Supercomputing 2020 Student Cluster Competition Reproducibility Challenge. We reproduce the single CPU-GPU performance experiments and the strong scaling experiments and compare the results to the original paper. Although we are not able to use the exact HPC system from the original paper, we use similar hardware and configurations to recreate the original environment in Microsoft Azure Cloud services. We were not able to demonstrate exact performance characteristics from the original paper due to hardware limitations in our environment, but we are able to reproduce similar performance trends for both single-node and scaling experiments. Digital artifacts from these experiments are available at: 10.5281/zenodo.5598108.