WarpX Accelerated Nodes Parallel Computing Paper
DOI10.5281/zenodo.4277941Zenodo4277941MaRDI QIDQ6716910FDOQ6716910
Dataset published at Zenodo repository.
Jean-Luc Vay, Luca Fideli, Eloise Yang, Kevin Gott, John Bell, A. Almgren, Olga Shapoval, David Grote, Weiqun Zhang, Neil Zaïm, Revathi Jambunathan, Axel Huebl, Remi Lehe, Henri Vincenti, Michael Rowan, L. D. Amorim, Lixin Ge, Yinjian Zhao, Maxence Thévenet, Mark Hogan, Cho Ng, Edoardo Zoni, Andrew Myers
Publication date: 31 October 2020
Copyright license: Creative Commons Attribution 4.0 International
This dataset contains the inputs, outputs, job submission scripts, and executables used to create the Figures in Porting WarpX to GPU-accelerated platforms by A. Myers et. al, submitted to Parallel Computing as part of the ECP Special Issue on Transitioning to Accelerated nodes. These results were obtained using the October, 2020 release tags of WarpX and AMReX, available on Github here: https://github.com/ECP-WarpX/WarpX and here: https://github.com/AMReX-Codes/amrex The following module files were loaded on Summit: 1) hsi/5.0.2.p5 2) xalt/1.2.0 3) lsf-tools/2.0 4) darshan-runtime/3.1.7 5) DefApps 6) cuda/10.1.243 7) gcc/6.4.0 8) spectrum-mpi/10.3.1.2-20200121 To use nsight-compute for the roofline plots, we also loaded: nsight-compute/2020.1.2 Manifest: BinScan: contains material used to make Figure 1. To generate the figure, use the Jupyter notebook called bin_size.ipynb. StrongScaling: contains material used to make Figure 5. To generate the figure, use the Jupyter notebook called strong_scaling.ipynb. WeakScalingCPU: contains material used to make Figure 4. To generate the figure, use the Jupyter notebook called weak_scaling.ipynb. WeakScalingGPU: contains material used to make Figure 5. To generate the figure, use the Jupyter notebook called weak_scaling.ipynb. Roofline: contains material used to make the roofline plots (Figures 2 and 3). This includes output generated using nsight-compute with WarpX and python scripts for processing and plotting these output files. These scripts and methodology originally come from Charlene Yang at NERSC. The file script.sh was used to generate the profiler output
This page was built for dataset: WarpX Accelerated Nodes Parallel Computing Paper
