Introduce multiple iterations for each test to eliminate interference introduced by the environment.
Skip the first result due to warmup, calculate the average of the subsequent ones as the final result.