Off-chain preprocessing and mandatory on-chain computations

[storojs72]

One additional way to reduce the gas cost of on-chain verification is moving some parts of verification logic to the form of preprocessing, which client can execute off-chain and submit results to on-chain verifier along with the proof / public parameters. Those results can be utilised by the verifier and thereby final gas cost can be reduced.

Can we identify what kind of computations in the whole e2e Nova verification are not so-to-speak "security sensitive" (may be there is better term for that) and can be moved to preprocessing phase performed off-chain, and which ones should be performed mandatory on-chain?

The data loader, that we currently use for uploading proof / public parameters data to the node's global storage seems to be very natural place for doing those off-chain preprocessing.

Seems that G1/G2 points decompression is a good candidate for being included into off-chain preprocessing.