Avoid excessive serialization/deserialization by using swapmap instead of backed map
Erase excessive legacy abstraction layers
Small optimizations for the vector clock algorithm
Performance test for epochs 15000-15100:
Before: 135.3s
After: 79.3
(1.7x speedup)
This metric includes only event processing (saveAndProcessEvent) without block processing in offline mode. Let me know if you get inconsistent results with this one
The optimization depends on number of validators. The more validators, the larger the improvement
Replace vecmt with vecclock:
Performance test for epochs 15000-15100: Before: 135.3s After: 79.3 (1.7x speedup)
This metric includes only event processing (
saveAndProcessEvent) without block processing in offline mode. Let me know if you get inconsistent results with this one The optimization depends on number of validators. The more validators, the larger the improvement