Chendi Li (ML+System)

I am currently a PhD student at the University of Utah, and supervised by Prof. P. (Saday) Sadayappan. I got my master degree from the State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences (CARCH, ICT, CAS), and supervised by Prof. Yunquan Zhang. Please refer to my short resume. If you have any questions, feel free to contact me by email.

Research Interests

Publications

[IEEE ISPA 2021] Chendi Li, Haipeng Jia, Hang Cao, et al. AutoTSMM: An Auto-tuning Framework for Building High-Performance Tall-and-Skinny Matrix-Matrix Multiplication on CPUs(presentation)(link,arxiv,ieee)

[IEEE ICPADS 2021] Jianyu Yao, Boqian Shi, Chunyang Xiang, Haipeng Jia, Chendi Li, et al. IAAT: An Input-Aware Adaptive Tuning framework for Small GEMM(link,ieee)

[IEEE HPCC 2021] Tun Chen, Haipeng Jia, Zhihao Li, Chendi Li, Yunquan Zhang. A Transpose-free Three-dimensional FFT Algorithm on ARM CPUs(link)

[CCF HPC China 2020] Chendi Li, Guangting Zhang, Haipeng Jia. Fast Computation of Elementary Functions on ARM Platforms(in Chinese)

Research projects

Designed AutoTSMM, which is used to build high-performance tall-and-skinny matrix multiplication on mainstream CPUs. AutoTSMM can speed up convolution layers in real-world deep learning applications, and the performance is competitive with Intel OneMKL and outperforms all conventional GEMM implementations. This work was published in IEEE ISPA 2021.

Optimized pre-pack matrix-matrix multiplication and triangular solve with multiple right-hand-sides(TRSM) on ARMv8 and X86 platforms. OpenBLAS is one of the most famous open-source BLAS libraries.

Launched the project and investigated JIT tools for small GEMM. IAAT is a template-driven just-in-time(JIT) small GEMM framework targeting CPUs. This work was accepted by IEEE ICPADS 2021.

Enhanced the math functions by manipulating IEEE 754 floating points. OpenVML is a vector mathematical library. It achieves an outstanding performance improvement compared to C standard library and ARMPL. This work was accepted by HPC China 2020.

Optimized small-scale FFT, and contributed to multi-threading and 2D-FFT. AutoFFT is a template-based FFT codes auto-generation framework that contributes to many Chinese vendors’ libraries. This work was published in SC’19, TPDS’20, and was accepted by HPCC’21.

Awards & Honors

2021: First-Class Scholarship of Chinese Academy of Sciences
2020: Second-class scholarship of Chinese Academy of Sciences
2019: Third-Class Scholarship of Chinese Academy of Sciences
2019: Outstanding intern in PerfXLab
2015: Outstanding volunteer
2015: Collegiate programming contest first prize

Invited Talks

ISPA’21: AutoTSMM: An Auto-tuning Framework for Building High-Performance Tall-and-Skinny Matrix-Matrix Multiplication on CPUs (IEEE ISPA 2021, Virtual Conference)(Youtube)

HPC-CHINA’20: OpenVML: Fast Computation of Elementary Functions on ARM Platforms(CCF HPC CHINA 2020, Virtual Conference)

Services

Student Volunteer at HPC CHINA 2021
Student Volunteer at HPC CHINA 2020
Student Volunteer at HPC CHINA 2019