Currently, every time a Knowledge Interaction is activated the ReasonerProcessor is recreated and part of this process is the costly graph pattern matching algorithm. Under certain circumstances (i.e. when the network is stable), the ReasonerProcessor can be reused which improves the performance considerably.
Doing a quick and dirty and naive implementation of this in the InteractionProcessorImpl where we reuse the ReasonerProcessor object when encountering the same KnowledgeInteraction object shows that for the TestPostReactPerformance test where the same knowledge interaction is repeated 1000 times it improved the performance per interaction from 9.5ms to 4.5ms. So, almost twice as fast.
So, this is an obvious place to further improve the performance of the knowledge engine when the network is (quite) stable.
Currently, every time a Knowledge Interaction is activated the
ReasonerProcessoris recreated and part of this process is the costly graph pattern matching algorithm. Under certain circumstances (i.e. when the network is stable), theReasonerProcessorcan be reused which improves the performance considerably.Doing a quick and dirty and naive implementation of this in the
InteractionProcessorImplwhere we reuse theReasonerProcessorobject when encountering the sameKnowledgeInteractionobject shows that for theTestPostReactPerformancetest where the same knowledge interaction is repeated 1000 times it improved the performance per interaction from9.5ms to4.5ms. So, almost twice as fast.So, this is an obvious place to further improve the performance of the knowledge engine when the network is (quite) stable.