Public Attributes
int	reps_min
int	reps_max
double	cl
double	eps
double	time_max_rep
double	zero_speed
double	time_max_kill

Detailed Description

Precision of measurement. Used as an input argument of benchmark functions. To provide reliable results, the computation experiments in each benchmark are repeated either fixed or variable number of times. This data structure allows the user to control the accuracy and efficiency of benchmarking.

Assigning to reps_min and reps_max the same values results in the fixed number of repetitions of the computation operation, with the cl and eps arguments being ignored (this allows the user to control the efficiency of benchmarking).
If reps_min < reps_max, the experiments are repeated until a confidence interval, fupermod_point::ci, found with the confidence level, cl = $Pr(|\bar T - \mu| < ci)$ , satisfies $\displaystyle\frac{ci}{\bar T} <$ eps, or the number of repetitions reaches its maximum, reps_max (this allows the user to control the accuracy of benchmarking).
If total benchmarking time exceeds time_max_rep then no more repetitions are attempted.
If zero_speed > 0, pthreads are used which allows the benchmark to be killed after a time interval x, where x = min(complexity(d)/zero_speed, time_max_kill).