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
Currently, each node holds a uint64 for the regCount and a uint16 for a maxDepth. The first one captures the number of registers s and the second one captures the maximum depth under that subtree. The register counts can be computed and kept on trie level instead of each node keeping that info. This is possible by knowing which values are adding new and which registers are being updated.
By keeping this data on trie level we can save a lot of memory space. Note that the checkpointer has to also be updated to not lose the counts after recovery.
The text was updated successfully, but these errors were encountered:
I can get started on this whenever you want, I think it is pretty straight-forward and has a lot of similarity with #1747. Those two could be tackled first.
Currently, adding an attribute to the trie that gets updated during update calls will require some changes in the design of the update function and even perhaps of the trie itself if we want to attach update to it as a method.
Currently, each node holds a uint64 for the regCount and a uint16 for a maxDepth. The first one captures the number of registers s and the second one captures the maximum depth under that subtree. The register counts can be computed and kept on trie level instead of each node keeping that info. This is possible by knowing which values are adding new and which registers are being updated.
By keeping this data on trie level we can save a lot of memory space. Note that the checkpointer has to also be updated to not lose the counts after recovery.
The text was updated successfully, but these errors were encountered: