Abstract: We present HPVM2FPGA, a framework that enables hardware-agnostic programming of FPGAs by coupling compiler optimization techniques with Design Space Exploration (DSE).
By using a suitable compiler Intermediate Representation designed for heterogeneous parallel systems, called HPVM, suitable compiler optimizations, and a state-of-the-art DSE framework (HyperMapper), we created an extensible flow that automatically generates high-performing code for FPGAs. HPVM2FPGA speedups between ~2x and ~33x compared to unoptimized baselines, and can meet the performance of hand-tuned HLS code when the hand tuning can be expressed as compositions of our supported transformations.
Bio: Adel is a PhD Candidate in Computer Science advised by Prof. Vikram Adve and Prof. Rob Rutenbar. He obtained his Bachelor of Engineering and Master of Engineering degrees in Electrical and Computer Engineering from the American University of Beirut in 2012 and 2015 respectively. Adel's research interests lie in both Computer Architecture and Compilers. His current research focuses on enabling hardware-agnostic programming of FPGAs by leveraging compiler techniques and design space exploration.