Closed ratankaliani closed 5 days ago
SP1 Performance Test Results
Branch: ratan/return-empty-bytes-groth16 Commit: 64753139 Author: ratankaliani
program | cycles | execute (mHz) | core (kHZ) | compress (KHz) | time | success |
---|---|---|---|---|---|---|
fibonacci | 11291 | 0.18 | 2.78 | 0.46 | 25s | ✅ |
ssz-withdrawals | 2757356 | 17.48 | 128.61 | 35.09 | 1m19s | ✅ |
tendermint | 12593597 | 6.71 | 267.42 | 98.68 | 2m9s | ✅ |
Motivation
Inconsistent behavior in
bytes
between mock PLONK and Groth16 proofs. This caused an issue inop-succinct
.In the future, especially for PR's from external folks, we should have simple tests added in the relevant file to test behavior. This prevents issues such as this which lead to inconsistent API behavior.
Reversion introduced by https://github.com/succinctlabs/sp1/pull/1685. Likely because base
dev
branch was not up to date.Solution
Return an empty byte array from
.bytes()
if it's a mock Groth16 proof.Add testing to confirm that
.bytes()
returns the correct output in the future.Note: Did not test
SP1Proof::Compressed
because constructing an emptyReduceProof
was difficult.PR Checklist