The directory contains 6 csv files, each containing either performance or energy measurements.

The files names is defined as follow:

"problem-size_type-measure"

problem-size indicates either:
- small: problem size used for the measurements 
- large: problem size used for the measurements
type-measure indicates either:
- energy: containing the sum of the energy consumption
- time_med: containing the median of the performance
- time_sum: containing the sum of the performance

Each file provides a list of parallel regions as well as a set of configurations: it describes the measurement of each region on the target configuration.
The configurations name is defined as:

compile-option_nb-threads_nb-numa-nodes-threads_nb-numa-nodes-pages_thread-mapping_SMT_prefetching_data-mapping

for instance: -O3_16_2_2_scatter_0_5_loc
compile-option: -O3, we systematically used -O3 
nb-threads: 16
nb-numa-nodes-threads: 2
nb-numa-nodes-pages: 2 (the same as nb-numa-nodes-threads in our experiments)
thread-mapping: scatter (or contiguous)
SMT: 0 (refering to last in the paper) 
prefetching: 5 : means that 2 prefetchers are disabled
data-mapping: locality
