You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current approach tries to take the given number of measurements,
of at least the given time, minimizing the median absolute deviation,
a robust measure of the variability.
It then reports the mean of the measurements and the maximal deviation from the mean.
As the performance can improve dramatically (with JIT and other effects),
it is necessary to have a dynamic number of iterations to respect the
minimal given time. The warm-up phase is then essentially spent in finding
an appropriate number of iterations.
Comparison will be done using the visual test using confidence intervals
and an approximation for the execution time ratio will be given.