On 17 Apr 00, at 2:12, Siegmar Szlavik wrote:
> This is a good option and I use it very often, especially on slower
> machines. What I am missing is some 'screensaver-functionality' (i.e.
> prime95 stops when there is some user interactivity). I would find it very
> usefull if this can be added to furture versions of prime95 (or even
> better to the nt service) George?
Hey, even the screensaver steals cycles when the system is active -
it needs to check that the system still is active!
If what you mean is, tap the screensaver API and make prime95 execute
only if the screensaver is/would be active, then fair enough, I'll
support that - provided it's an option. (This is rather similar to
mprime running under linux starting & stopping depending on loadavg)
>
> yes, it "should", but I noticed, that on some machines prime95/ntprime
> gets out of control after a few days and the system is no longer useable.
? (see below)
> The program seems to get all the cpu time, allthough it runs with the
> default (lowest) priority. Sometimes it helps to reboot the machine once
> or twice a day, but just only sometimes :( Any idea, what could be the
> reason?
Right. It is _perfectly normal_ for a compute-bound process to get
essentially all the CPU time, providing that nothing with a higher
priority is in a computable state.
The questions that need to be asked are:
(1) is the system really grinding to a halt for some reason? In this
case, it sounds as though this may be the case. In other words,
_something_ is stopping interactive programs getting to a computable
state.
(2) if so, what is causing this? (Bearing in mind that Prime95 seems
to be blissfully ignorant!)
Given that this state is taking time to emerge, I would suspect that
the cause is a memory leak reducing the physical memory available to
interactive processes to a point where the system performance is
badly hit. Prime95 has a nice chunk of memory from when it started &
doesn't change this allocation much, so it just chunters on. (Though
its own performance will start to deteriorate as page/swap thrashing
becomes established).
Now Prime95 seems to run "forever" on an otherwise idle system, so I
think we can rule out Prime95 as being the source of the memory leak.
Note that, if I'm right on this, the system would eventually slow
down badly without Prime95 running at all - though it might well take
longer for the effects to become obvious.
The best diagnostic for this would be to use the Win 9x System
Monitor tool (Start/Programs/Accessories/System Tools) - there is a
similar tool in NT, and a much improved version if you happen to have
the appropriate Resource Kit installed - to monitor the memory usage
of the system services (the programs which run all the time Windows
is up), and also of the system free memory (which you have to deduce
from the disk cache size). After a few hours monitoring, you should
be able to spot the culprit.
Microsoft have in the past released patches for some Windows
components and also for some of their applications which have
corrected memory leaks. If a patch is not available, but the leak is
in another component which is not essential (e.g. the notorious
FindFast service which Office installs for you), then you can also
fix the problem by not running that component.
If you have inadvertently installed cracker tools like BackOrifice
etc. then you'd also expect deteriorating performance. It would
definitely be worthwhile checking your systems - even if you decide
to give up running Prime95.
I'd certainly check out the patch level of your Windows components
and whatever applications you are using and bring them up to date.
This is worth doing anyway, e.g. Service Pack 6a for Win NT v4.0
corrects a number of known problems with security (as well as general
fixes) which can make your systems vulnerable to various denial-of-
service attacks.
Basically, what I'm saying is that your systems seem to be getting
into a state they shouldn't, and I don't think Prime95/NTPrime is
playing a part in this. You're seeing Prime95/NTPrime taking a big
slice of CPU time on a system that appears to be choked up, but
that's because it's about the only thing wnating CPU time left in a
runnable state, not a result of the system being overwhelmed by its
demands. In other words, Prime95/NTPrime is being blamed for a
problem which would exist anyway. The problem needs to be fixed, not
what is only a symptom of the problem, and very probably an
incorrectly interpreted symptom at that.
Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
