Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
portfolio
HPC-Bench
A benchmark of 114 HPC workloads across 16 computational motifs evaluating LLM-driven optimization in serial CPU, OpenMP and CUDA. It combines reproducible builds, correctness checks and the Fast@k / Speedup@k metrics. The study finds that reliable optimization becomes harder as required speedup increases, particularly on CUDA.
HPC-Performance-AI
An ongoing framework for performance prediction and optimization in HPC applications. The open benchmark infrastructure spans GPU microbenchmarks, proxy/mini-applications and production scientific codes, with standardized build, execution, timing and validation workflows. Performance prediction and optimization components remain active research goals.
LLM-Based HPC Optimization Study
An earlier study of 24 HPC programs across nine computational domains, comparing LLM-driven serial and parallel optimization against compiler-based tools. Its correctness and performance findings motivated the expanded HPC-Bench evaluation.
TileBench
A controlled benchmark comparing 45 operators in Triton and cuTile on NVIDIA B200. Matched semantics, autotuning, NCU/PTX diagnosis and LLM-generated kernels expose workload-dependent compiler and hardware performance differences.
publications
A Novel Deep Reinforcement Learning Strategy in Financial Portfolio Management
Cite
B. Cui, R. Sun, J. Su. (2022). "A Novel Deep Reinforcement Learning Strategy in Financial Portfolio Management." In 2022 7th International Conference on Big Data Analytics (ICBDA), pp. 341–348.
Do Large Language Models Understand Performance Optimization?
Public preprint on understanding LLM capabilities and limitations in performance optimization.
Comprehensive Evaluation of LLMs in HPC Code Performance Optimization
Evaluated 24 HPC programs from nine domains, comparing LLM-based serial and parallel optimization with traditional performance tools.
HPC-Bench: A Comprehensive Benchmark for High Performance Computing Codes
A correctness-gated benchmark of 114 HPC workloads across 16 computational motifs, covering serial CPU, OpenMP and CUDA code optimization.
TileBench: A Controlled Benchmark for Performance Evaluation and Bottleneck Diagnosis of Tile-Based Programming Models
An evaluation of 45 GPU operators under matched semantics in Triton and cuTile, with autotuning, compiler inspection and profiling on NVIDIA B200.
Estimating Regional Snow Water Equivalent Using a Physics-Informed Graph Convolution Network Model
Physics-informed graph convolutional models for spatial snow water equivalent estimation across the western United States.
talks
teaching
Teaching experience 1
Undergraduate course, University 1, Department, 2014
This is a description of a teaching experience. You can use markdown like any other post.
Teaching experience 2
Workshop, University 1, Department, 2015
This is a description of a teaching experience. You can use markdown like any other post.
