Two ideas. If you run the benchmark (runme_xeon64 for example), you will see GFlops increasing with problem size. The size of the problem is a control of processor load. By limiting number of threads via environment variable OMP_NUM_THREADS you can decrease the load.
Thanks for help, I'm still looking for some advice though.
Actually what I'm aiming for is putting a steady load on my cpu over a period of time (around 15 minutes) and I'd like to be able to control the amount of that load (to keep it for example at 50% and at another time around 75%).
Any ideas on how could I achieve that using this benchmark?
Hi Krzysztof S.
about 15 minutes, then how about try some big(enough) problem sizes
or run the one big workload,and run ./runme_xeon64 ; ./runme_xeon64 several times together, thus the cpu workload can keep about 50%