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
For now, we only have 1 example of a proven sector for the production 1GiB sector size, which means certain extrinsics' weights (declare_faults, declare_faults_recovered, terminate_sectors) aren't accurately scaling with the number of sectors.
When we have test data for more sectors (or figure out a good way to test it without them), the benchmarks need to be updated with n: Linear<1, N> parameters and re-run.
As the code is written right now we'd need to run separate PoRep proofs of the 1GiB sector data, which each takes 20 minutes, and right now DECLARATIONS_MAX == 3000 and MAX_TERMINATIONS_PER_CALL == 32 which means about 3 solid years of proving.
Once we figure out what would be reasonable values for those constants and how much would even fit in a block (using non-scaled weights as an initial estimate), then we can decide on a good limit to run it to and scale it linearly from there.