Video Decoding & AI Acceleration
Energy-efficient video processing and machine learning acceleration. Chongqing University · Sep 2017 – Jul 2020
Research group: Chongqing University · Sep 2017 – Jul 2020
Researched energy-efficient video processing and machine learning acceleration during my graduate studies.
Contributions:
- Parallelized H.264 decoding at the task level and applied dynamic voltage and frequency scaling, improving decoding speed by 27–32% while reducing energy consumption by 23–25% across 720p–2160p video (ICPADS 2018).
- Contributed to HolyLight, a nanophotonic accelerator for data-center CNN workloads, achieving 13× higher throughput per watt than a ReRAM baseline (DATE 2019).
Related publications: ICPADS 2018 · DATE 2019