Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Jiahui Xu, Emmet Murphy, Lana Josipovic

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).

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

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).
Paper Structure (10 sections, 3 figures)

This paper contains 10 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Building an MLIR-Based HLS Compiler in an Academic Setting
  4. Limitations of MLIR for Open-Source HLS Projects and Case Studies in Dynamatic
  5. MLIR Value is Not a Fit for Modeling Graph Edges
  6. Are Block Arguments Convenient for HLS Conversion?
  7. Have We Solved the Software Segmentation Problem?
  8. Do We Have a Good HLS C Frontend that Encourages People to Build New HLS Tools?
  9. Beyond HLS
  10. Conclusion

Figures (3)

  • Figure 1: Annotating edge information in MLIR is tricky.
  • Figure 2: Dynamatic aims to translate software IR (left) to dataflow circuit (right).
  • Figure 3: Integrating two MLIR projects is challenging.