09/19/2023 CHARM Demo at ESWEEK2023 ESSC!

My students Jinming Zhuang and Zhuoping Yang and I gave a remote demo “CHARM: Composing Heterogeneous AcceleRators for End-to-end Deep Learning Inference on Versal ACAP Architecture” on THE EMBEDDED SYSTEMS RESEARCH SOFTWARE COMPETITION (ESSC) in ESWEEK 2023. Thanks to the ESSC Chairs Ganapati Bhat and Biresh Kumar Joardar, ESWEEK 2023 Chairs X. Sharon Hu and Alain Girault for offering this great opportunity for us to showcase the open-source system software and artifacts and enrich open-source tools for the EDA community! In this demo, we show how to use CHARM, our open-source tool to compile deep learning inference on AMD Versal ACAP architecture (AI ASIC + FPGA + CPU). This repo has received 76 stars since we released it in Feb.

You are welcome to check our paper for this demo:

1. FPGA2023 CHARM: Composing Heterogeneous Accelerators for Matrix Multiply on Versal ACAP Architecture. It’s the Top 1 Downloaded paper in the FPGA 2023 proceeding.

2. DAC2023 High Performance, Low Power Matrix Multiply Design on ACAP: from Architecture, Design Challenges and DSE Perspectives

Thanks to our collaborators in this project, Prof. Jingtong Hu, Prof. Alex Jones, Prof. Deming Chen, Prof. Jason Cong, student collaborators Hanchen Ye, Jason Lau, and our AMD collaborators Stephen Neuendorffer, Jack Lo, and Kristof Denolf!

LinkedIn