The standalone calibration tool is designed to calibrate the model that is used for physical hosts. These scripts aim to work
for gathering calibration data that is suitable for the energy modeller.

The first stage is to start the monitoring.

./gather_calibration_data.sh

The second is to apply a load. The parameter to be passed is how many stream multiprocessors are on the gpu.

./gpu-stress.sh <count_of_sm_proccessors>

once the load has finished, the data gathering process can be stopped.

./stop_gathering.sh

Files with the pattern: <Hostname>-<card_name>-<gpu_index>-stats.csv will be produced. Such as: ns51.bullx-GPU-Tesla K20Xm-0-stats.csv 

The file that demonstrates load being applied can be moved into the folder with the energy modeller and renamed.

It should be given the same name as the graphics card (<card_name>.csv)  i.e. "GPU-Tesla K20Xm.csv"