This folder contains all the files for the second exercise. The goal of this exercise is to compare performance of three math libraries available on HPC: MKL, OpenBLAS and BLIS. The comparison is performed focusing on the level 3 BLAS function called gemm.
gemm_parallel.cis a standard gemm code, where 3 matrices A,B,C are allocated, A and B are filled and the BLAS routine calculates the matrix-matrix product C=A*B, using parallelization via OpenMP.Makefileis the makefile used to compile the code.*.share the scripts used to run the code on the cluster.$NODE_sbatch.shfiles are used for the scalability over the matrix size, while$NODE_cores.shfiles are used for the scalability over the number of cores.data/directory contains the data produced by the code: it is organized in subdirectories, one for each node used. Each subdirectory contains the data for the scalability over the matrix size, and the scalability over the number of cores.
After deciding which node to use and what kind of scalability to test, open the corresponding script and change the location of your BLIS installation. Then, run the script with sbatch $NODE_$SCALABILITY.sh. The data will be saved in the corresponding subdirectory of data/. The sbatch files already compile the code, so there is no need to do it manually.