You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
speaking of space used for permutations, I just noticed that we could probably halve it in the Software Heritage graph by using CompactArray instead of naive arrays, as we have 34 billion nodes (2^35).
Do you want me to try and see the impact on LLP performance?
The text was updated successfully, but these errors were encountered:
Well, the only difference should be in the generation and inversion of the permutation. The problem is that CompactArray in the atomic version does not have very strong concurrency guarantees, so you wouldn't be able to use parallel sorting when generating it.
speaking of space used for permutations, I just noticed that we could probably halve it in the Software Heritage graph by using CompactArray instead of naive arrays, as we have 34 billion nodes (2^35).
Do you want me to try and see the impact on LLP performance?
The text was updated successfully, but these errors were encountered: