11/13/2023
By Martin Margala
The Francis College of Engineering, Department of Electrical & Computer Engineering, invites you to attend a doctoral defense by Shachivaman Khakilkar on “Open-source CAD tooling for FPGAs."
Candidate Name: Shachivaman Khakilkar
Degree: Doctoral
Defense Date: Nov. 27, 2023
Time: 10 a.m. to noon
Location: ECE Department Conference Room. Those interested in attending should contact the student shachivaman_khadilkar@student.uml.edu and committee advisor martin_margala@uml.edu.
Advisor: Martin Margala, Department of Electrical & Computer Engineering, University of Massachusetts, Lowell
Committee Members:
- Yan Luo, Department of Electrical & Computer Engineering, University of Massachusetts Lowell
- Vinod Vokkarane, Department of Electrical & Computer Engineering, University of Massachusetts Lowell
- Ahmed Sanaullah, RedHat
Brief Abstract:
Field Programmable Gate Arrays (FPGA) have gained traction due to their reprogrammability, power efficiency and high performance. Commercial state-of-the-art Computer Aided Design (CAD) software needed to map computations to the FPGA is typically tailored to a particular FPGA device/family. These commercial CAD tools are closed-source and cannot be customized. Open-source alternatives often lack significantly in performance and Quality of Results (QoR) compared to the highly tuned commercial CAD tools, making the former impractical for use in non-trivial applications.
This research is aimed at bridging the hardware quality gap between open-source and vendor FPGA CAD software. We compare the result quality between state-of-the-art vendor CAD tools and equivalent open-source tools and quantify the gap between them. We also propose a generic input-aware optimization framework that determines the best algorithm policies for a given digital circuit and device architecture. We evaluate this framework using the packing step of the compilation flow and present the results in this thesis. This work also provides results on autotuning FPGA design and optimization parameters using reinforcement learning techniques.
Bio:
Shachi is a fourth-year Ph.D. candidate supervised by Prof. Martin Margala. She completed her Master of Science degree in Computer Engineering from Boston University. Her Master’s studies focused on analog and RF IC design. She worked at Boston University’s CIDAR lab before joining Prof. Martin Margala’s research group at the University of Massachusetts Lowell. In the early stages of her PhD, she started working as a research intern at Red Hat, focusing on open-source CAD tooling for FPGAs. Her doctoral research area is optimization techniques for open-source FPGA CAD software.