Is It a Good Idea to Build an HLS Tool on Top of MLIR? Experience from Building the Dynamatic HLS Compiler

TL;DR

This paper discusses the potential and challenges of using MLIR as the foundation for High-Level Synthesis (HLS) tools, sharing insights from building the Dynamatic compiler.

Contribution

It provides an empirical evaluation of MLIR's suitability for HLS by sharing practical experiences from developing the Dynamatic compiler.

Findings

MLIR offers both opportunities and challenges for HLS development.

Building Dynamatic revealed specific limitations of MLIR for HLS tasks.

The paper provides recommendations for future HLS tool development using MLIR.

Abstract

When the MLIR project was first introduced, it promised to address the issues that the HLS community had with the LLVM project. But is this really the case, and is MLIR the "right"/"best" compiler infrastructure for HLS? We here share our experiences based on the development of Dynamatic (github.com/EPFL-LAP/dynamatic).