GPU-Accelerating Process Simulation Performance using NVIDIA’s cuDSS Sparse Linear Systems Solver

, Senior System Software Engineer, NVIDIA
, Senior Fellow, Honeywell Connected Enterprise
NVIDIA has developed a new large-scale solver, cuDSS, for sparse linear systems that uses GPU computations for the matrix factorization and solution. This solver was integrated into the UniSim EO platform using the UniSim AXB sparse linear algebra interface, which enables sparse linear algebra packages to be easily integrated. Honeywell has a set of matrices that were generated from UniSim EO process model applications, which includes models of refining, petrochemical, and chemical process units, as well as models of processes for greenhouse gas recovery. Honeywell’s Dotaxb sparse linear equation solver was used as the basis for comparison with cuDSS and the computing platform was an Azure server with an NVIDIA A100 Tensor Core GPU. Learn how the cuDSS solver showed performance improvement of 2-6x over Honeywell’s sparse linear equation solver for the test matrices, which in the past has outperformed other solvers (including multicore solvers) that it was tested against.
活动: GTC 24
日期: March 2024
话题: Accelerated Computing Libraries
NVIDIA 技术: CUDA
行业: Energy
级别: 中级技术
语言: 英语
所在地: