The original logic creates a HashedReferenceList when the list is very small (which is when the optimization would have least impact). This generates a lot of allocations when creating tiny Pools and also seems unintended. Now, the hashed implementation is used for large Pools (where it is likely to improve performance), and the list is used as-is if small enough.
embeddedt
commented
7 months ago
The original logic creates a
HashedReferenceListwhen the list is very small (which is when the optimization would have least impact). This generates a lot of allocations when creating tinyPools and also seems unintended. Now, the hashed implementation is used for large Pools (where it is likely to improve performance), and the list is used as-is if small enough.