As #89 said, fury io.fury.Fury#serializeJavaObject(java.io.OutputStream, java.lang.Object) may save a copy inside, which may affect the benchmark result.
This won't make very much difference since serialization is much more time consuming than the copy of ~250 bytes, but to make the benchmark result more solid, this PR workaround it by copying the data manually .
As #89 said, fury
io.fury.Fury#serializeJavaObject(java.io.OutputStream, java.lang.Object)may save a copy inside, which may affect the benchmark result.This won't make very much difference since serialization is much more time consuming than the copy of
~250bytes, but to make the benchmark result more solid, this PR workaround it by copying the data manually .