migrating to mediawiki
hendrik at pooq.com
Sun Oct 23 00:30:58 PDT 2005
On Sat, Oct 22, 2005 at 08:24:38PM -0400, Far? wrote:
> On 19/10/05, Hendrik Boom <hendrik at pooq.com> wrote:
> > I've been working on a threaded-code interpreter that may satisfy
> > some of the needs of the TUENS project. It's not FORTH, it's strongly
> > typed, it's garbage-collected, and it's barely reaching the pre-alpha stage.
> > It is designed so that it *could* have its innards replaced by something
> > that generates efficient, optimised machine-code without doing much to
> > the programs run on it.
> Have you taken a look at the CAM that serves as the basis of CAML?
> It's a mostly stack virtual machine. See Xavier Leroy's article about
> "The Zinc Experiment", in which he reimplemented CAML using a portable
> bytecode. CAML was originally implemented in Le-Lisp. Xleroy
> bootstrapped it into using his bytecode interpreter, yielding
> caml-light, which evolved into a wordcode interpreter, caml-sl
> (special light), which became OCaml, and nowadays also has a native
> code compiler.
Thanks. I'll have to look that up and see if it is relevant.
> > Unlike the discussions that took place years and years ago on one
> > of the TUNES forums, it type-marks the integers instead of the pointers.
> Remarkably, so does the OCaml VM.
Yes. It turns out that one uses pointers mich more within the
thresded code interpreter than integers. So keeping the type-marks off
them is a decided convenience. I discovered this the hard way.
More information about the TUNES