Closed Sword-Smith closed 5 months ago
map
function calls are another 7.5 % it seems. I wonder if we should also look into that.
One step could be to write a merkle_root snippet for a list of XFEs as opposed to a list of digests and similar specialized snippets.
The latest profile of a verifier execution shows that
zip
is responsible for 7.2 % of both the opstack table rows and processor table rows. I'm not sure we can get rid of all of the use ofzip
without changing theFRI
verifier return type. But let's see how far we can get. Also consider whether we want to change this interface.Profile of verifier of inner padded height $2^{20}$, expansion factor 8