Closed leruaa closed 1 week ago
SP1 Performance Test Results
Branch: aurelien/gro-8-sp1-output-for-local-proving-with-rust_loginfo-is-messy Commit: 941356b0 Author: leruaa
program | cycles | execute (mHz) | core (kHZ) | compress (KHz) | time | success |
---|---|---|---|---|---|---|
fibonacci | 11291 | 0.16 | 2.90 | 0.47 | 24s | ✅ |
ssz-withdrawals | 2757356 | 16.24 | 126.50 | 34.54 | 1m20s | ✅ |
tendermint | 12593597 | 6.70 | 268.39 | 100.29 | 2m7s | ✅ |
Motivation
Execute and prove_core end up printing out the same stuff twice, like cycle counts and syscall invocations. It's confusing for users to figure out if there should be a difference between these two outputs
A bunch of stuff about trace area for every table gets printed out, which users don't care about and is just confusing.
This refers more to the cuda prover I think: We could change the default to be RUST_LOG=sp1_cuda=off,info or something similarbi
Solution
PR Checklist