Title | ||
---|---|---|
Characterization of Data Generating Neural Network Applications on x86 CPU Architecture |
Abstract | ||
---|---|---|
This paper analyzes the performance of two contemporary data-generating neural network-based workloads, Neural Style Transfer and Super Resolution GAN run on x86 hardware architecture. In understanding the impact of data-readiness, we find how certain layers benefit from forced data warming-up. In examining bandwidth utilization of these layers, we identify several memory-bound layers as not necessarily being bandwidthbound hinting at the feasibility of prefetch-based solutions for improved performance. We also observe layers with specific kernel sizes performing poorly because of their unoptimized library kernel implementation. Based on our findings, we suggest directions for removing these performance bottlenecks by utilizing available bandwidth margins ≥ 90% and realizing convolution operations through vector-based functional units with a scope of at least 20x more such software-to-hardware mappings than existing implementation. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1109/ISPASS48437.2020.00026 | 2020 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) |
Keywords | DocType | ISBN |
software-to-hardware mappings,convolution operations,data generating neural network applications,super resolution GAN,prefetch-based solutions,memory-bound layers,bandwidth utilization,forced data,data-readiness,x86 hardware architecture,Neural Style Transfer,contemporary data-generating neural network-based workloads,x86 CPU architecture,vector-based functional units,unoptimized library kernel implementation | Conference | 978-1-7281-4799-4 |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Antara Ganguly | 1 | 0 | 0.34 |
Shankar Balachandran | 2 | 0 | 0.34 |
Anant Nori | 3 | 19 | 3.01 |
Virendra Singh | 4 | 218 | 40.22 |
Sreenivas Subramoney | 5 | 127 | 13.60 |