Misc,

Replication Data for: Distributing Higher-Dimensional Simulations Across Compute Systems: A Widely Distributed Combination Technique

.
Software, (2021)Related to: Distributing Higher-Dimensional Simulations Across Compute Systems: A Widely Distributed Combination Technique; Theresa Pollinger, Marcel Hurler, Michael Obersteiner, and Dirk Pflüger, 2021. In: 2021 IEEE/ACM International Workshop on Hierarchical Parallelism for Exascale Computing (HiPar), pp. 1-9. doi: 10.1109/HiPar54615.2021.00006.
DOI: 10.18419/darus-2123

Abstract

Scripts to run the weak, strong and remote strong scaling tests described in "Distributing Higher-Dimensional Simulations Across Compute Systems: A Widely Distributed Combination Technique", as well as measurements obtained on the Machines HAWK and our Institute's epyc machine. To reproduce, follow the build procedure detailed in the scripts repository , making sure that the flags passed in `DisCoTec/distributedcombigrid/examples/distributed_third_level/Makefile.template` match the ones passed during library compilation. Observe the folder structure for the different tests performed: To run a specific test case, the executable `combi_example` was linked into a folder, and the parameter file `ctparam` was provided. The files called `run.sh` are batch system submission scripts. `generate_test_folders_and_submit.sh` automates the process for a series of weak scaling runs, which can be adjusted by modifying the script. Time measurement outputs are given as `timers.json` files, where each process wrote a portion of the file at the end (can get very large for many processes and many time steps). The `plot_*` files in discotec-scaling-scripts help extract the relevant information from the `timers.json` files.

Tags

Users

  • @unibiblio

Comments and Reviews