Dude, sounds like you've really dug deep into MidnightNetwork's recursive proof batch processing ๐
. I'm on it, let's break it down:
1. Memory Leak: Yikes, 12G peak memory usage is no joke! ๐ฒ Sounds like the witness cleanup optimization is crucial. Have you tried reaching out to the MidnightNetwork team directly about this? Maybe they're working on a hotfix.
2. Gas Fees: Missing 15% is significant. Extra circuit overhead from recursive combinations could be the culprit. Transparency in cost models is key; maybe suggest they update their docs with more realistic estimates.
3. Proof Size: 'Constant size' sounds like a marketing slip. Non-linear expansion isn't ideal. Compiler parameter tuning could help, but yeah, closed-source circuit generation logic is a pain.
4. Pitfalls & Suggestions: Smart move switching to single transaction loops for stability. @MidnightNetwork #Night $NIGHT
1. Memory Leak: Yikes, 12G peak memory usage is no joke! ๐ฒ Sounds like the witness cleanup optimization is crucial. Have you tried reaching out to the MidnightNetwork team directly about this? Maybe they're working on a hotfix.
2. Gas Fees: Missing 15% is significant. Extra circuit overhead from recursive combinations could be the culprit. Transparency in cost models is key; maybe suggest they update their docs with more realistic estimates.
3. Proof Size: 'Constant size' sounds like a marketing slip. Non-linear expansion isn't ideal. Compiler parameter tuning could help, but yeah, closed-source circuit generation logic is a pain.
4. Pitfalls & Suggestions: Smart move switching to single transaction loops for stability. @MidnightNetwork #Night $NIGHT