EdgeLLM: A Highly Efficient CPU-FPGA Heterogeneous Edge Accelerator for Large Language Models

TL;DR

EdgeLLM introduces a CPU-FPGA heterogeneous accelerator that significantly improves the efficiency and throughput of deploying large language models on resource-constrained edge devices, addressing computation, memory, and deployment challenges.

Contribution

The paper presents a novel heterogeneous accelerator architecture, a universal data parallelism scheme, and an end-to-end compiler for efficient LLM deployment on edge devices.

Findings

Achieves 1.91x higher throughput than NVIDIA A100 GPU.

Attains 7.55x higher energy efficiency compared to GPU.

Outperforms state-of-the-art FPGA accelerators by 10-24% in performance.

Abstract

The rapid advancements in artificial intelligence (AI), particularly the Large Language Models (LLMs), have profoundly affected our daily work and communication forms. However, it is still a challenge to deploy LLMs on resource-constrained edge devices (such as robots), due to the intensive computation requirements, heavy memory access, diverse operator types and difficulties in compilation. In this work, we proposed EdgeLLM to address the above issues. Firstly, focusing on the computation, we designed mix-precision processing element array together with group systolic architecture, that can efficiently support both FP16*FP16 for the MHA block (Multi-Head Attention) and FP16*INT4 for the FFN layer (Feed-Forward Network). Meanwhile specific optimization on log-scale structured weight sparsity, has been used to further increase the efficiency. Secondly, to address the compilation and deployment issue, we analyzed the whole operators within LLM models and developed a universal data parallelism scheme, by which all of the input and output features maintain the same data shape, enabling to process different operators without any data rearrangement. Then we proposed an end-to-end compiler to map the whole LLM model on CPU-FPGA heterogeneous system (AMD Xilinx VCU128 FPGA). The accelerator achieves 1.91x higher throughput and 7.55x higher energy efficiency than the commercial GPU (NVIDIA A100-SXM4-80G). When compared with state-of-the-art FPGA accelerator of FlightLLM, it shows 10-24% better performance in terms of HBM bandwidth utilization, energy efficiency and LLM throughput.