一句话：提出任务并行的数据流编程框架 TAPA-CS，可自动划分并跨 FPGA 集群编译大规模设计，同时保证高频与高吞吐。
针对云规模多 die FPGA 缺乏高效编程工具的问题，本文提出 TAPA-CS：一个任务并行的数据流编程框架，可自动将大规模设计划分并编译到 FPGA 集群。它开源并让用户利用近乎“无限”的加速器资源、HBM 与片上内存；能自动划分设计以实现拥塞控制、资源均衡及通信计算重叠；并将粗粒度布局规划与 FPGA 间/内互连流水化结合以保证高频。实验在 100Gbps 以太网连接的多 FPGA 测试床上评估脉动阵列 CNN、PageRank、模板计算与 KNN 等设计。相比 Vitis HLS 单 FPGA 基线，2、3、4 FPGA 设计平均加速 2.1×、3.2×、4.4×，频率提升 11%–116%。
Despite the increasing adoption of FPGAs in compute clouds, there remains a significant gap in programming tools and abstractions which can leverage network-connected, cloud-scale, multi-die FPGAs to generate accelerators with high frequency and throughput. We propose TAPA-CS, a task-parallel dataflow programming framework which automatically partitions and compiles a large design across a cluster of FPGAs while achieving high frequency and throughput. TAPA-CS has three main contributions. First, it is an open-source framework which allows users to leverage virtually "unlimited" accelerator fabric, high-bandwidth memory (HBM), and on-chip memory. Second, given as input a large design, TAPA-CS automatically partitions the design to map to multiple FPGAs, while ensuring congestion control, resource balancing, and overlapping of communication and computation. Third, TAPA-CS couples coarse-grained floor-planning with interconnect pipelining at the inter- and intra-FPGA levels to ensure high frequency. FPGAs in our multi-FPGA testbed communicate through a high-speed 100Gbps Ethernet infrastructure. We have evaluated the performance of TAPA-CS on designs, including systolic-array based CNNs, graph processing workloads such as page rank, stencil applications, and KNN. On average, the 2-, 3-, and 4-FPGA designs are 2.1×, 3.2×, and 4.4× faster than the single FPGA baselines generated through Vitis HLS. TAPA-CS also achieves a frequency improvement between 11%-116% compared with Vitis HLS.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{abs-2311-10189,
title = {TAPA-CS: Enabling Scalable Accelerator Design on Distributed HBM-FPGAs},
author = {Neha Prakriya and Yuze Chi and Suhail Basalama and Linghao Song and Jason Cong},
booktitle = {ASPLOS},
year = {2024},
doi = {10.1145/3620666.3651347}
}