Research
Research Interests
I am a researcher in the field of high performance computing with a focus on data compression and sparse matrix storage. My research interests include: HPC, data compression, sparse matrix storage, sparse algorithms, CUDA programming, and performance portability. I am currently working on the development of high performance compression algorithms for the FZ Project and the IVSparse Library. I have also had experience as an undergraduate research intern at Sandia National Laboratories working on the E3SM atmospheric model, HOMME (High-Order Method Modeling Environment).
Projects
FZ (Recently Joined Researcher)
The FZ Project is a lossy error-bound lossy compression framework funded by the NSF. These compressors see use in many fields such as AI, quantum simulation, scientific computing, and instrumentation.
I am currently working on the development team of the IU FZ Team under the advisement of Dr. Fengguang Song. Thus far I’ve worked on profiling the performance of the cuSZ family of compressors and analyzing the perofrmance of key kernels on different Nvidia GPUs. Going forward I will be working on developing further high performance compression kernels for the FZ team that are performance portable across different architectures using the Kokkos library.
IVSparse (Primary Co-Contributor)
The IVSparse Library is a C++ sparse matrix library optimized for the compression of sparse data in cases where non-zero values are highly redundant. IVSparse supports three main compression formats.
- CSC: Compressed Sparse Column - The standard sparse matrix format that stores the column indices of non-zero values and the values themselves.
- VCSC: Value Compressed Sparse Column - A format that takes advantage of the redundancy in the values of the matrix by storing unique values and their counts.
- ~2.25x fold compression over CSC (for redundant data)
- IVCSC: Indexed Value Compressed Sparse Column - A format that further compresses the index storage of VCSC by using positive-delta encoding and then bytepacking indices for each unique value in a column.
- ~7.5x fold compression over CSC (for redundant data)
Publications
- [BigData ‘24] Seth Wolfgang, Skyler Ruiter, Marc Tunnell, Timothy Triche Jr, Erin Carrier, Zachary DeBruine. “Value-Compressed Sparse Column (VCSC): Sparse Matrix Storage for Single-cell Omics Data.” 2024 IEEE International Conference on Big Data (BigData). Washington, DC, USA. December 15-18. arxiv
Talks and Pictures
