Move vkey window precomputation off-chain
Closed this issue · 0 comments
schradeyannik commented
Current state:
- on-chain precomputation for each public input with number of public inputs
$n$ : ~$n * 32 * 255 * 64$ bytes
advantages of this system:
- guaranteed correctness of the precomputes values
- linking using accounts (-> does not affect binary size)
disadvantages:
- optimized but still computation intensive (multiple thousand tx required)
- very tedious to perform computations per proof updating (especially in devnet)
Possible alternatives:
- account + manual direct binary upload (tx size limit will bring us into a similar ballpark for required work though)
- vkey as binary file + linking in binary directly (possibly with 7 instead of 8 bits window size then)