Closed robert3005 closed 6 days ago
I have found this while spot checking profiles - in q19 computing true range was 30% of all our compute time
I can maybe believe it, but the benchmark run doesn't agree with you?
It’s overall 5% better or am I looking at wrong numbers?
The q19 row didn't show an improvement which seems suspicious given the profile
Yeah so I have run this on top of #1466 on my laptop and that was an improvement. I imagine our ci machines would produce a different profile. Could also be that this is bigger improvement with #1466 that it is standalone
True range is too expensive of a metric to calculate given current apis