Adjust plot.bin to implement sequential s_bucket reading

the audit index is stored in the query, and then all sectors are looped to read the data pointed to by the audit index. Can we adjust plot.bin after plotting so that s_bucket can finish reading in order?

1 Like


You may misunderstand the algorithm a bit. S-buckets are already stored sequentially, so when you’re auditing sector, you read the whole thing in one go quickly.

However, audit indices are different for each sector, so even if we store s-buckets of the same offset of each sector together it doesn’t help it any way and it causes a lot of issues with data retrievability and farm resizing.

You might be interested in watching Dilithium: Consensus v2 Implementation Walkthrough - April 25, 2023 where I go into details about how things are processed, layout of the data on disk (there were checksums added later, but otherwise it is still accurate) and why.

Let me know if I misunderstood you.