Closed hargoniX closed 3 months ago
According to: https://drops.dagstuhl.de/storage/00lipics/lipics-vol271-sat2023/LIPIcs.SAT.2023.21/LIPIcs.SAT.2023.21.pdf the trimming is basically a depth first search followed by a renaming to keep indizes sequential. We might just be able to reimplement this in Lean and still cut down significantly on the size of the LRAT proofs.
https://github.com/arminbiere/lrat-trim can be useful for us in two ways:
Eval.Popcount
:c trimmed 91256 added clauses in original proof to 56188 clauses 62%
) running this as a post processing step could speed up verification further.