Our tests on the Deneb branch are noticeably slower, and it appears that this is caused by the blob commitment/proof generation in tests that uses the block generator to generate random blobs.
Below is a flamegraph capture by @michaelsproul on one of the slow tests, which shows that a significant amount of time is spent on computing kzg commitments / proofs:
We could potentially speed up these tests by pre-generating a bunch of (blobs,commitment,proof) tuples to usein the beacon chain tests (idea suggested by @pawanjay176).
jimmygchen
commented
1 year ago
Description
Our tests on the Deneb branch are noticeably slower, and it appears that this is caused by the blob commitment/proof generation in tests that uses the block generator to generate random blobs.
Below is a flamegraph capture by @michaelsproul on one of the slow tests, which shows that a significant amount of time is spent on computing kzg commitments / proofs:
We could potentially speed up these tests by pre-generating a bunch of
(blobs,commitment,proof)tuples to usein the beacon chain tests (idea suggested by @pawanjay176).