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
From discussions with @draenog, it would be safer to calculate benchmarks by averaging a significantly longer duration (minimum ~1s), rather than our current approach of averaging many shorter iterations.
This would be straightforward is we just wanted to calculate the mean for the entire loop, but finer details may be more complicated, e.g. do we need to be careful about how it ends up being optimised, and is there a straightforward way to time the allocation of arrays repeatedly etc.?
The text was updated successfully, but these errors were encountered:
From discussions with @draenog, it would be safer to calculate benchmarks by averaging a significantly longer duration (minimum ~1s), rather than our current approach of averaging many shorter iterations.
This would be straightforward is we just wanted to calculate the mean for the entire loop, but finer details may be more complicated, e.g. do we need to be careful about how it ends up being optimised, and is there a straightforward way to time the allocation of arrays repeatedly etc.?
The text was updated successfully, but these errors were encountered: