arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2304.12474cs.ARcs.LG
收藏

使用 FPGA 进行高性能计算的设计优化

Design optimization for high-performance computing using FPGA

Drexel University(德雷塞尔大学);Omer Halisdemir University(奥默·哈利斯德米尔大学)

Murat Isik, Kayode Inadagbo, Hakan Aktas

AI总结:

本文通过硬件改进、Xilinx Ultra RAM 和编译器策略优化 Tensil AI 开源 FPGA 加速器,在 CIFAR ResNet20 上实现高能效推理。

摘要:

诸如现场可编程门阵列(FPGA)之类的可重构架构,凭借其在灵活性、性能和功耗效率方面的独特结合,已被用于加速多个领域的计算。然而,FPGA 尚未在高性能计算中得到广泛应用,主要原因在于其编程复杂性以及性能优化方面的困难。为了深入了解 FPGA 在高性能计算中的使用,本文针对在 CIFAR 上训练的 ResNet20,对 Tensil AI 的开源推理加速器进行优化,以实现最高性能。本文展示了改进硬件设计、使用 Xilinx Ultra RAM 以及采用先进编译器策略如何能够提升推理性能。我们还证明,运行 CIFAR 测试数据集时,从原始 32 位浮点数向下舍入只会造成极小的精度下降。平台中的异构计算模型使我们能够在使用 CIFAR 训练的 ResNet20 上实现 293.58 帧每秒(FPS)的帧率和 90% 的准确率。实验结果表明,所提出的加速器在 100 MHz 下实现了 21.12 Giga-Operations Per Second(GOP/s)的吞吐量,片上功耗为 5.21 W。与现成器件以及近期最先进实现的比较结果表明,所提出的加速器在能效方面具有明显优势。

英文摘要:

Reconfigurable architectures like Field Programmable Gate Arrays (FPGAs) have been used for accelerating computations in several domains because of their unique combination of flexibility, performance, and power efficiency. However, FPGAs have not been widely used for high-performance computing, primarily because of their programming complexity and difficulties in optimizing performance. We optimize Tensil AI's open-source inference accelerator for maximum performance using ResNet20 trained on CIFAR in this paper in order to gain insight into the use of FPGAs for high-performance computing. In this paper, we show how improving hardware design, using Xilinx Ultra RAM, and using advanced compiler strategies can lead to improved inference performance. We also demonstrate that running the CIFAR test data set shows very little accuracy drop when rounding down from the original 32-bit floating point. The heterogeneous computing model in our platform allows us to achieve a frame rate of 293.58 frames per second (FPS) and a %90 accuracy on a ResNet20 trained using CIFAR. The experimental results show that the proposed accelerator achieves a throughput of 21.12 Giga-Operations Per Second (GOP/s) with a 5.21 W on-chip power consumption at 100 MHz. The comparison results with off-the-shelf devices and recent state-of-the-art implementations illustrate that the proposed accelerator has obvious advantages in terms of energy efficiency.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑