Pipeline Balancing for Integrated Mapping in High Performance Spatial Programmable Architecture

Pengyu Liu,Zihan Zhang,Chen Yin,Liyan Chen,Jianfei Jiang,Qin Wang,Zhigang Mao,Naifeng Jing

DOI: https://doi.org/10.1109/fpl60245.2023.00024

2023-01-01

Abstract:Recently, spatial programmable architectures have gained increasing popularity owing to their performance and programmability, while the achievable performance is highly related to how the operators are mapped onto a number of processing elements (PEs) in the spatial architectures. In this paper, we first identify that in the spatial mapping process, the pipeline balancing problem is essential by affecting the spatial initial interval (SpII). Hence, we formulate the SpII for the first time in spatial architecture. The quantitative formulation enables an integrated mapping algorithm which combines operator placement, operand routing and pipeline balancing at the same time. In addition, to reduce the balancing hardware cost, we propose a bridge-buffer structure to facilitate operand routing and buffering on demand. To reduce the mapping searching space, we propose three optimization techniques to trade off solution quality, mapping time and hardware overhead. The experiment results show that the proposed integrated mapping algorithm can reduce the SpII by 42.3%, which in turn improves the throughput and algorithm runtime up to 1.74× and 3.08× over the state-of-the-art heuristic spatial mapping.