A High-Density Data Path Implementation Fitting for HTC Applications
Wenming Li,Liang Zhang,Xiaochun Ye,Da Wang,Hao Zhang,Zihou Wang,Dongrui Fan
DOI: https://doi.org/10.1109/igcc.2015.7393734
2015-01-01
Abstract:High Throughput Computing (HTC) applications become the new loadings with the rapid rising of web services. In HTC applications, as we observed, a significant proportion of memory accesses are in small granularity, such as 1B or 2B. However, the link width is usually designed as 128 bits or even larger to achieve high throughput in traditional NoCs. The entire bandwidth is occupied no matter how large the fiit is. Therefore, using traditional NoCs for HTC applications will lead to the waste of bandwidth. In this paper, to address the above-mentioned problem, we proposed High-Density NoC (HD-NoC). In HD-NoC, traditional link is split into several narrow channels, such as 8 or 16 bits. If the slice is 16 bits wide, there will be 8 or more separately self-governed small channels running simultaneously in one direction. Cooperating with our Greedy Transfer Mechanism (GTM), flits in the same direction can be transferred parallel, which will alleviate the congestion and improve effective utilization of bandwidth. Experiments show that for HTC applications, our proposed HD-NoC improves throughput rate by 22.2% in average and 32.4% for Grep application with little extra hardware resources. The HD-NoC is also able to improve throughput rate by 13.5% for traditional SPLASH-2 benchmarks.