Responsibilities include:

• Qemu profiling on a branch landing scheme
• Proposed two extension on branch landing scheme
• Designed a workflow to evaluate the proposed extension

• Supervised by Xuehai Qian, Purdue University
• Analyzed resent defense strategy like InvisiSpec and CleanupSpec;
• Designed a buffer-based Undo approach to mitigate the transient speculation flaw.
• Extended the current memory coherence protocol to support the merging and purging requests in our design.

• Supervised by Zhibin Yu and Xuehai Qian
• Used genetic algorithm to automatically generate extremely high power consumption.
• Modified gpgpusim simulator to trace the access pattern for gpgpu simulations.

• Supervised by Zhiru Zhang
• Implemented both the hardcore and softcore of the BNN network on an FPGA hardware.
• Coded for the interface to connect the Rocket chip softcore with the BNN accelerator.
• Used High Level Synthesis tool Stratus to utilize limited resources to implement the project.

• Supervised by Shouyi Yin
• Used deep learning algorithms to analyze human behavior while driving a vehicle.
• Used the deep learning platform “tensorflow” to solve traditional problems, e.g. MNIST classification.
• Investigated the mechanism behind deep learning algorithms.