Open MLIR Meeting 8/11/2022: HeteroCL, an MLIR-based Intermediate Representation for Accelerator Design with Decoupled Customizations

This Thursday, August 11th (9am California Time, 16:00 UTC), Hongzheng Chen & Niansong Zhang (Cornell) will present HeteroCL: an MILR-based Intermediate Representation for Accelerator Design with Decoupled Customizations.

With the pursuit of improving compute performance under strict power constraints, there is an increasing need for deploying applications to heterogeneous hardware architectures with accelerators, such as GPUs and FPGAs. However, although these heterogeneous computing platforms are becoming widely available, they are very difficult to program especially with FPGAs. As a result, the use of such platforms has been limited to a small subset of programmers with specialized hardware knowledge.

To tackle this challenge, we introduce HeteroCL, a programming infrastructure comprised of a Python-based domain-specific language (DSL) and a compilation flow. The HeteroCL DSL provides a clean programming abstraction that decouples algorithm specification from three important types of hardware customization in compute, data types, and memory architectures. HeteroCL can further capture the interdependence among these different customization techniques, allowing programmers to explore various performance/area/accuracy trade-offs in a systematic and productive manner. In addition, our framework currently provides two advanced domain-specific optimizations with stencil analysis and systolic array generation, which produce highly efficient microarchitectures for accelerating popular workloads from image processing and deep learning domains.

Project page: GitHub - cornell-zhang/heterocl: HeteroCL: A Multi-Paradigm Programming Infrastructure for Software-Defined Heterogeneous Computing

Zoom Meeting Link

Meeting ID: 851 5109 0498
Passcode: 828404


Thanks everyone for attending our presentation today. For more information about our work, please check our hcl-mlir branch. We are still doing some code cleaning and refactoring in our dialect, and will open-source the MLIR part soon. Please keep an eye on that.

We have actually submitted several patches to improve the Python binding of the MLIR dialect. Later, we will think about how to decompose and integrate our efforts into the upstream MLIR in order to benefit a broader range of communities. We’re welcome to discuss more about this.

If you have further questions or have collaboration plans, please feel free to contact me (Hongzheng) or Niansong. Thanks!

Here are slides and recording for today.