Offload ECDSA verification to a new coprocessor #241
Labels
crate: evm_arithmetization
Anything related to the evm_arithmetization crate.
performance
Performance improvement related changes
Milestone
Once #227 and #172 are implemented, the
TrieData
segment overhead on memory should be light enough that we can process large transaction batches at once (possibly up to an entire block).Given that each transaction requires a signature verification, which currently accounts for more than 26k CPU cycles, this could represent significant savings. Current L1 blocks orbit around 80 and 280 txns, which could correspond to 2M+ / 7M+ CPU cycles for a whole block.
For additional perspective, native asset transfers take around 32.5k CPU cycles for actual txn processing, and ERC20 token transfers around 51k, for which
ECRECOVER
represents 81% and 52% of the overall CPU overhead respectively.The text was updated successfully, but these errors were encountered: