I read in the release notes of the version 4.1.0 a mention of HashLink Immix and wasn’t able to find out what is the difference with HashLink, except that it looks faster.
current benchs
JVM-HL/C(LMMIX)-HXCPP…
it seem the new hl gc bring very fast performance.
I mentioned this at the end of my HaxeUp 2019 talk – I am developing a new GC for HashLink as part of my Master’s thesis.
It looks promising in the benchmarks, although it is still a work in progress! It needs to be debugged a lot and refactored in a way that the new GC can be simply enabled with a compilation flag (rather than having to checkout a completely separate fork of HL).
The principal GC method behind it is Immix (2008), just like for hxcpp. However, hxcpp has grown a lot since “just Immix”.
Stay tuned. 
5 Likes