We use the complete set of parallel workloads provided in
PARSEC-2 benchmark suit [3]. For each benchmark, we have
simulated region of interest and the first 10% of the simulated
instructions are considered as warm up period and statistics are
gathered at the end of program execution.