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
Since we don't pre-allocate, the vec starts with size 1, and often requires re-allocating. Would we start with a default size of, say 5?
(don't ask me why 5, just a magic number!)
Opening an issue to track this improvement. The change can improve current throughput, benchmark by 10-20% (as they all need ~5 attributes.)
The text was updated successfully, but these errors were encountered:
I think we discussed it earlier - If we plan to start with default size, we can also evaluate using smallvec, to allocate this initial capacity on stack. tokio-tracing is doing something similar for the metrics attributes.
Since we don't pre-allocate, the vec starts with size 1, and often requires re-allocating. Would we start with a default size of, say 5?
(don't ask me why 5, just a magic number!)
Opening an issue to track this improvement. The change can improve current throughput, benchmark by 10-20% (as they all need ~5 attributes.)
The text was updated successfully, but these errors were encountered: