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
The current API feels rather leaky at the moment, and is currently trading lowered memory usage for extra access costs on every access of the Merkle tree.
Instead by duplicating the nodes (using copy) and using an internal slab allocator, we could improve overall performance, clean up the API, at the cost of negligable extra memory cost.
Expected Behavior
Tell us how the feature should work. Explain the feature largely in terms of
examples.
Drawbacks
We'd be relying on internalized allocation, which might bite us in future for FFI. On the other hand: if we only require allocations to happen for data that crosses the FFI boundary, we don't need to worry here as this is very much internal. I think at least.
Worst case we can expose an additional API once we get to an FFI layer.
Rationale and alternatives
Alternatively we could leave the current API as-is, but it feels a bit messy, and is weird to use.
Unresolved Questions
None.
The text was updated successfully, but these errors were encountered:
Feature Request
Summary
Change
to
Motivation
The current API feels rather leaky at the moment, and is currently trading lowered memory usage for extra access costs on every access of the Merkle tree.
Instead by duplicating the nodes (using
copy
) and using an internalslab allocator
, we could improve overall performance, clean up the API, at the cost of negligable extra memory cost.Expected Behavior
Tell us how the feature should work. Explain the feature largely in terms of
examples.
Drawbacks
We'd be relying on internalized allocation, which might bite us in future for FFI. On the other hand: if we only require allocations to happen for data that crosses the FFI boundary, we don't need to worry here as this is very much internal. I think at least.
Worst case we can expose an additional API once we get to an FFI layer.
Rationale and alternatives
Alternatively we could leave the current API as-is, but it feels a bit messy, and is weird to use.
Unresolved Questions
None.
The text was updated successfully, but these errors were encountered: