This is amazing, particularly the much smaller memory use, which I assumes comes because structured types can be used by value in Go (vs. the all heap approach of Java), isn't it? I assume the parallelization of the front-end is explicit in the source code (haven't looked at the source so far); what Go concepts did you use for it?
You could even consider re-using the Go compiler IR and backend for native, statically linked binaries, instead of generating JVM IR.
You could even consider re-using the Go compiler IR and backend for native, statically linked binaries, instead of generating JVM IR.