[DGD] DGD/MP 1.0

[Felix A. Croes]

Kirktis <kirktis at gmail.com> wrote:

> One question I just realized that I don't believe I've seen come up
> yet, is the subject of virtual multiprocessing, as presented in
> Intel's Hyperthreading model. Is this something that DGD/MP will be
> able to take advantage of, or would that be futile in the grand scheme
> of things?

In the grand scheme of things, hyperthreading was a temporary hack in
the leadup to multi-core processors.

Hyperthreading is something that DGD can "take advantage of" in the
sense that it will happily accept a fake 2nd CPU which is presented
by the OS as a real one.  In this case, a virtual advantage can work
out as a decided disadvantage.  In the benchmarks I've seen,
hyperthreading results in a speedup of about 5%.  Unless your mudlib
guarantees that less than 5% of all execution rounds will be
rescheduled -- which would be quite remarkable -- the overall result
will be a slowdown.

Regards,
Dworkin