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
Comment by Constantine creator on how he made constantine library so fast:
Nim has better macro and compile-time evaluation capabilities so I implemented an assembly codegenerator in Nim macro.
Most Rust finite fields libraries return their result instead of in-place mutation. This is slow if the compiler doesn't manage to do return value optimization.
algorithm differences
Parallelization difference: I parallelize at 3 different levels of the MSM. Rust lib parallelize at only one-level. Also Rayon doesn't support futures and you can use it for significant latency hiding in bucket reductions.
The underlying parallel library in general is underestimated and I improve performance by up to 50% with just a better barrier algorithm here for CPU with 16+ cores: mratsim/constantine#227 (comment)
The text was updated successfully, but these errors were encountered:
Comment by Constantine creator on how he made constantine library so fast:
The underlying parallel library in general is underestimated and I improve performance by up to 50% with just a better barrier algorithm here for CPU with 16+ cores: mratsim/constantine#227 (comment)
The text was updated successfully, but these errors were encountered: