This artifact's aim is to reproduce results reported in our paper. The results are obtained on two systems located at OLCF. This artifact contains a folder for each system as follows: summit/ install-slate-on-summit.sh run-slate-on-summit.sh slate/ crusher/ install-slate-on-crusher.sh run-slate-on-crusher.sh slate/ Each folder has an installation script to install SLATE. The installation scripts can be run as follows: cd summit/ cd slate/ source ../install-slate-on-summit.sh The installation script for Crusher install-slate-on-crusher.sh can be run in the same way. Note that these installation scripts must be sourced since they load system modules and change environment variables. These modules and environment variables are also required by the experiments. The following commands are used to reproduce the gemm and potrf results on Summit: cd summit/slate/ bsub ../run-slate-on-summit.sh The following commands are used to reproduce the gemm and potrf results on Crusher: cd crusher/slate/ bash ../run-slate-on-crusher.sh Note that the account numbers in run-slate-on-....sh scripts must be set accordingly. The largest number in the gflop/s column in the result output is reported as the performance of the kernel operation on a corresponding number of nodes of the system.