On Sunday 22 February 2009 11:26:14 Philippe Strauss wrote:
> Remembering other needs like a concurrent GC and the multicore GC project
> of the OSP 2008, does the idea of modifiying the GC of ocaml to make it a
> "pluggable" subsystem, with 3 runtimes you may choose from, the
> "traditional" GC, the "real-time, pool based", and the "concurrent" one
> sounds reasonable?
>
> Did this kind of need often pop-up in your mind?

Yes, although I have been considering what might be done with OCaml's 
bytecode. Presumably it conveys all of the necessary information in a more GC 
agnostic way already. Moreover, a JIT compiler built upon LLVM could easily 
outperform the existing OCaml compilers.

You may be interested in these benchmark results I gathered for different VMs:

  http://flyingfrogblog.blogspot.com/2009/01/mono-22.html

-- 
Dr Jon Harrop, Flying Frog Consultancy Ltd.
http://www.ffconsultancy.com/?e