Closed Nashtare closed 2 years ago
@baumbata, I believe you had looked at this about 2 months ago, did you find anything blocking?
I don't think there was anything blocking. I do have code written on a branch somewhere I could dust off. It just didn't get tested before I switched gears to working on the constraints for the composite AIR program.
Partially addressed with PR #18. Would need something like RocksDB to manage such a large tree with regular machines. Will consider reopening depending on the integration path when integrated to our framework.
We're currently using two different Tree depth constants for unit tests and general use mode, i.e. 4 or 16.
The actual expected state of a subnet would be rather 2^32, which would require to generate an empty tree directly through hardcoded nodes at each layer, to prevent tree creation time to blow-up. It would be also helpful to test performances in real conditions.
This would be directly implemented in https://github.com/Toposware/winterfell/tree/main/crypto/src/merkle/mod.rs with the other helper methods.