This is a tiny change I found while pruning my local branches, and I wouldn't want us to miss it.
The guard on atomic_batch is utilized a lot, which - in the current setup - also causes the guard on checkpoints used in the atomic_checkpoint method to be in force significantly longer than necessary.
This tiny change with no impact on logic fixes this, as confirmed by guard lifetime measurements in a --dev run; the maximum guard (on checkpoints) duration is reduced by ~75%.
ljedrz
commented
4 months ago
This is a tiny change I found while pruning my local branches, and I wouldn't want us to miss it.
The guard on
atomic_batchis utilized a lot, which - in the current setup - also causes the guard oncheckpointsused in theatomic_checkpointmethod to be in force significantly longer than necessary.This tiny change with no impact on logic fixes this, as confirmed by guard lifetime measurements in a
--devrun; the maximum guard (oncheckpoints) duration is reduced by ~75%.