Yafan Huang | University of Maryland

Selected Software

cuSZp: GPU lossy compression framework for floating-point HPC data and ML tensors.
LLMFI: Fault injector for studying error propagation in large language model inference.
lsCOMP: Fast GPU lossy/lossless compressor for light source data and unsigned integers.

Some interesting tools developed during my PhD. I will be happy if that helped you before.

Understanding LLVM Transformation Passes (in Chinese)
This is a Chinese book I wrote during my spare time as a Ph.D. student. The motivation behind this work stems from my belief that LLVM transformation passes are incredibly powerful tools, yet the existing documentation is often vague and lacks sufficient detail. In particular, this book aims to address that gap by providing: (1) An introduction to LLVM basics. (2) A quick overview of several fundamental compiler optimization techniques. (3) A detailed explanation of each LLVM (Version 17.0.0) transformation pass, including definitions, my interpretations, and illustrative code examples. I hope this book serves as a helpful resource for those looking to deepen their understanding of LLVM transformation passes. Enjoy reading!
cuSZp Documentation
I developed and currently maintain cuSZp, an ultra-fast GPU compression framework designed to tackle big data challenges with high-speed requirements. This is a detailed documentation for cuSZp explanation and usages. If you’re working with large-scale data and need efficient compression, please feel free to explore the GitHub repository and give it a try!

A Brief Introduction to LLVM
This document introduces basic concepts of the LLVM compiler infrastructure and provides a practical example of how to write a simple LLVM pass for performing static program analysis.
Using LLFI: LLVM IR Level Fault Injector
This document provides an overview of using LLFI, an LLVM-level fault injector developed by the University of British Columbia (UBC), to study the error resilience of programs.

I recevied my Ph.D. from Computer Science at the University of Iowa (2021 - 2026) advised by Guanpeng Li. I was also a visiting graduate student at Argonne National Laboratory, where I worked closely with Sheng Di and Franck Cappello. Before coming to UIowa, I earned my Master’s degree in Computer Science from Huazhong University of Science and Technology (2018 - 2021), and my Bachelor’s degree in Software Engineering from Hunan University (2014 - 2018).