Code Transpilation for Hardware Accelerators

TL;DR

This paper presents an integrated approach combining code translation and accelerator generation tools to simplify deploying legacy code on hardware accelerators, enhancing workflow efficiency and programmability.

Contribution

It introduces a novel integration of Metalift and Gemmini, enabling easier translation of legacy code to hardware accelerators and reducing development effort.

Findings

Streamlined workflow for running legacy code on accelerators

Reduced effort in adding new instructions to Gemmini

Potential for optimizing computational workloads

Abstract

DSLs and hardware accelerators have proven to be very effective in optimizing computationally expensive workloads. In this paper, we propose a solution to the challenge of manually rewriting legacy or unoptimized code in domain-specific languages and hardware accelerators. We introduce an approach that integrates two open-source tools: Metalift, a code translation framework, and Gemmini, a DNN accelerator generator. The integration of these two tools offers significant benefits, including simplified workflows for developers to run legacy code on Gemmini generated accelerators and a streamlined programming stack for Gemmini that reduces the effort required to add new instructions. This paper provides details on this integration and its potential to simplify and optimize computationally expensive workloads.