A Study of LS-DYNA® Implicit Running the Rolls-Royce® Large Representative Engine Model with Intel® Optane™ DC Persistent Memory Technology

In this paper we discuss Intel’s continued efforts to optimize the performance of LS-DYNA Implicit. We focus on the Rolls-Royce® Large Representative Engine Model (LREM), the largest implicit model known to Livermore Software Technology. Performance analysis indicated three opportunities for improvement: shared memory parallelization of the LS-GPart reordering code, optimization of the multifrontal linear solver, and usage of Intel® Optane™ DC Persistent Memory. We present results taken while running the LREM model with a tuned hybrid version of LS-DYNA R12.r144413 HYBRID_DP on 2nd Generation Intel® Xeon™ Platinum 8260L scalable processor (formerly Cascade Lake) cluster with Intel’s Optane persistent Memory. We depict the benefits of Intel Optane persistent Memory technology and discuss the techniques needed to optimize LS-DYNA.