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
When running benchmark on my Raptor Lake processor, I find the time of each iteration shifts between two values if I switch the console window to background.
This is probably due to the "smart" scheduling for P/E cores of Windows. I've also heard that thread affinity API getting ignored by the system in some situations.
It would also be interesting to see the results on E-cores. To be more general, there can be a configuration to set thread affinity for benchmarks. This can also help benchmarking for arbitrary core counts.
The text was updated successfully, but these errors were encountered:
When running benchmark on my Raptor Lake processor, I find the time of each iteration shifts between two values if I switch the console window to background.
This is probably due to the "smart" scheduling for P/E cores of Windows. I've also heard that thread affinity API getting ignored by the system in some situations.
It would also be interesting to see the results on E-cores. To be more general, there can be a configuration to set thread affinity for benchmarks. This can also help benchmarking for arbitrary core counts.
The text was updated successfully, but these errors were encountered: