Implementation and performance evaluation of NVDLA based deep learning accelerator hardware

• Main Author: Song, Tin Chen
• Other Authors: Lap-Pui Chau
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering::Microelectronics; Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: https://hdl.handle.net/10356/149624
• Institution: Nanyang Technological University

This report shows the steps needed for one to implement a deep Learning hardware accelerator based on NVIDIA DL accelerator NVDLA architecture on high-performance emulation. The simulation platform chosen was Firesim. NVDLA architecture is an industrial-grade opensource hardware accelerator project for Deep Learning inference acceleration. NVDLA not only provides full hardware design source file but also provides the corresponding software library to directly deploy DL networks that are trained and optimized using NVIDIA GPU based AI system. On top of the implemented NVDLA accelerator, evaluation on the performance of the hardware accelerator was done based on existing research papers.