I am currently a 1st-year Ph.D. student at MASA Lab, College of Computer Science and Technology, National University of Defense Technology (NUDT), under the supervision of Prof. Mei Wen and Dr. Yang Shi. My research interests primarily focused on AI accelerator architecture and compilation, sparse computing.

My previous work specialized in designing and optimizing parallel programs (including GEMM, SpMM) for AI accelerators, especially GPU Tensor Cores, achieving not-bad results in accelerating SpMM. I am also keenly interested in processor performance modeling, AI compilers, and architecture design, which I plan to pursue as my future research directions.

Interesting Projects

1. Low Precision GEMM on GPU Tensor Core.(Ongoing Updates!)
2. Unstructured SpMM on GPU Tensor Core.
3. Architecture-based Analytical Cost Model for Modeling, Auto-Tuning, HW Generation.(Next Stage)

Selected Publication

• SSpMM: Efficiently Scalable SpMM Kernels Across Multiple Generations of Tensor Cores. PDF. code.
• Releasing the Potential of Tensor Core for Unstructured SpMM using Tiled-CSR Format. PDF.