MARCA: Mamba Accelerator with ReConfigurable Architecture

TL;DR

MARCA introduces a reconfigurable accelerator architecture with novel reduction, nonlinear function, and buffer strategies, significantly enhancing speed and energy efficiency for deep learning workloads.

Contribution

The paper presents a novel reconfigurable architecture for the Mamba accelerator, including new reduction methods, nonlinear function units, and buffer management strategies.

Findings

Achieves up to 463.22× speedup over CPU.

Achieves up to 9761.42× energy efficiency over GPU.

Demonstrates effective reconfigurable nonlinear function execution.

Abstract

We propose a Mamba accelerator with reconfigurable architecture, MARCA.We propose three novel approaches in this paper. (1) Reduction alternative PE array architecture for both linear and element-wise operations. For linear operations, the reduction tree connected to PE arrays is enabled and executes the reduction operation. For element-wise operations, the reduction tree is disabled and the output bypasses. (2) Reusable nonlinear function unit based on the reconfigurable PE. We decompose the exponential function into element-wise operations and a shift operation by a fast biased exponential algorithm, and the activation function (SiLU) into a range detection and element-wise operations by a piecewise approximation algorithm. Thus, the reconfigurable PEs are reused to execute nonlinear functions with negligible accuracy loss.(3) Intra-operation and inter-operation buffer management strategy. We propose intra-operation buffer management strategy to maximize input data sharing for linear operations within operations, and inter-operation strategy for element-wise operations between operations. We conduct extensive experiments on Mamba model families with different sizes.MARCA achieves up to 463.22$\times$/11.66$\times$ speedup and up to 9761.42$\times$/242.52$\times$ energy efficiency compared to Intel Xeon 8358P CPU and NVIDIA Tesla A100 GPU implementations, respectively.