A Flexible Template for Edge Generative AI with High-Accuracy Accelerated Softmax & GELU

TL;DR

This paper presents a specialized hardware accelerator, SoftEx, for fast and energy-efficient softmax and GELU computations in Transformer-based generative AI models, boosting performance and efficiency.

Contribution

Introduction of SoftEx, a novel accelerator for softmax and GELU, integrated into a heterogeneous RISC-V based cluster, achieving significant speedups and energy savings.

Findings

SoftEx achieves up to 10.8x speedup for softmax.

SoftEx reduces energy consumption by up to 10.8x.

End-to-end ViT inference throughput increases by 1.58x.

Abstract

Transformer-based generative Artificial Intelligence (GenAI) models achieve remarkable results in a wide range of fields, including natural language processing, computer vision, and audio processing. However, this comes at the cost of increased complexity and the need of sophisticated non-linearities such as softmax and GELU. Even if Transformers are computationally dominated by matrix multiplications (MatMul), these non-linearities can become a performance bottleneck, especially if dedicated hardware is used to accelerate MatMul operators. In this work, we introduce a GenAI BFloat16 Transformer acceleration template based on a heterogeneous tightly-coupled cluster containing 256KiB of shared SRAM, 8 general-purpose RISC-V cores, a 24x8 systolic array MatMul accelerator, and a novel accelerator for Transformer softmax and GELU non-linearities: SoftEx. SoftEx introduces an approximate exponentiation algorithm balancing efficiency (121x speedup over glibc's implementation) with accuracy (mean relative error of 0.14%). In 12nm technology, SoftEx occupies 0.039 mm$^2$, only 3.22% of the cluster, which achieves an operating frequency of 1.12 GHz. Compared to optimized software running on the RISC-V cores, SoftEx achieves significant improvements, accelerating softmax and GELU computations by up to 10.8x and 5.11x, respectively, while reducing their energy consumption by up to 10.8x and 5.29x. These enhancements translate into a 1.58x increase in throughput (310 GOPS at 0.8V) and a 1.42x improvement in energy efficiency (1.34 TOPS/W at 0.55V) on end-to-end ViT inference workloads.