[RC5] Re: Needing testers (Mac OS X/G5)

Slawek sgp at telsatgp.com.pl
Tue Aug 10 17:16:26 EDT 2004


Hello!
In message to "D.net Discussion" <rc5 at lists.distributed.net> sent Mon, 9 Aug
2004 19:34:57 +0200 you wrote:


 JM> On 9 aug 2004, at 18:57, Slawek wrote:

 SP>> Ohh, so G5 is another processor with the problem similar to Pentium 4?
 SP>> I thought this was something specific to HyperThreading only.
 SP>>
 SP>> D.net benchmark works in normal priority - just as your other
 SP>> applications/daemons.
 SP>> If benchmark is slowed down by d.net background cruncher - your other
 SP>> processes are likely to be slowed down as well.

 JM> It has nothing to do with hyperthreading or something similar on the
 JM> G5. The "problem" is simply that Mac OS X' lowest priority level is not
 JM> the same as e.g. Linux' idle priority. There is no priority level on
 JM> Mac OS X where a process will only get the idle cycles. So even if you
 JM> run dnetc at priority 19 under Mac OS X, it'll slow down all other
 JM> applications somewhat (though other applications will get a larger
 JM> percentage of the available cycles).

Ok, thanks for the clarification.
I can understand it can be fine from OS's point of view and system's
administrator's point of view, but it looks like this is a real problem for
distributed.net - it causes computers running distributed.net client to work
slower.

There are obviously situations which will always work slower because of
distributed.net client.
For example:
1. background threads (zeroing freed memory, garbage collection) need to
fight for CPU with distributed.net
2. threads requiring much cache memory (cache would be taken by d.net client
whenever application's thread goes idle waiting for HDD or network or
something)
3. systems which throttle CPU clock on overheating (which are designed to do
that when CPU is used in 100%)
4. systems where running multiple CPUs simultanously causes each of the
processors to work a little bit slower then it works when being the only one
working at the moment (logical processors with Hyper Threading bring this to
extreme)

Unfortunatelly it looks like we can do just about nothing about those
problems. It could be clearly stated on download page at most.

-- 
Slawek Piotrowski




More information about the rc5 mailing list