Christie Alappat awarded third prize at ISC 2023 Best Poster Awards
ISC High Performance 2023, “The Event for High Performance Computing, Machine Learning, Data Analytics, and Quantum Computing,” took place from May 21-25 In Hamburg. Part of the vast program is a research poster session, which was particularly competitive this year. We are all the more happy to announce that our PhD student Christie L. Alappat was awarded the third prize in the ISC Best research Poster awards. Congratulations!
Christie’s poster, titled “RACE: Speeding Up Sparse Iterative Solvers Using Cache-Blocked Matrix Power Kernels,” is concerned with the optimization of one of the prevalent kernels in scientific computing: sparse matrix-vector multiplication (SpMV). This seemingly simple operation is responsible for using up a significant part of the compute resources in HPC centers. SpMV with large matrices is limited by memory access bandwidth or latency. However, if multiple back-to-back SpMVs with the same matrix have to be applied in sequence to a vector, cache blocking for the matrix data becomes possible. Christie’s method is based on RACE, the Recursive Algebraic Coloring Engine, and can achieve speedups of 2x-4x on modern Intel or AMD x86 multicore CPUs. The poster demonstrates the method and shows its application on iterative methods such as s-step Krylov solvers, polynomial preconditioners, and algebraic multigrid (AMG) preconditioners.
Download the full poster here: ISC_23_CA.pdf
We could not find any entry with the given search term 8660.