[RC5] Lousy Mac 2.8007-458 performance

Michael Feiri michael.feiri at mfeiri.lake.de
Sun Mar 26 00:14:47 EST 2000


on 21.03.2000 2:55 Uhr, Elliot Robinson at galaxy4000 at earthlink.net wrote:

> I asked this question a while back.  The answer I got was something like
> rc5des 2.7105.xxx wasn't reporting an accurate speed, which I don't find
> very believable.

But its true. The old Meggs and DiCamillo both used Ticks to measure time.
But Ticks seem to drift on fast machines. We have evaluated quite a number
of timing methods and now we use the very accurate PCI timer which in theory
has a nanosecond resolution rather than milliseconds as in Ticks.

> Still, the last Meggs client gets a higher
> benchmark/when idle speed than the 458 dnetc build.  On the other hand if
> you have a G4, you have nothing to complain about... :)

Better benchmark results might be due to the inaccurate timer as noted
above. But you are of course right that the current clients ability to use
spare cycles is severely limited. The current client just behaves as a
normal application while Meggs and DiCamillo clients used some tricky
dynamic aggressiveness regulation routines. The new client will soon get
similar features again.

Ciao, Michael d:-)

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list