To compile the benchmarks you need to specifiy your library paths

make all EIGENROOT=... BLAZEROOT=... LIBXSMMROOT... FASTORROOT=...

Single and double precision benchmarks are run one at a time you need to define make RUN_SINGLE=-DRUN_SINGLE for single precision benchmarks. The default is double precision

To run the benchmarks

make run >> benchmark_results_double.txt

Note that you will have to run the benchmark a few times to get a sense of where each library stands.

To plot the the results, you need python with numpy and matplotlib

python benchmark_plot.py