[Skimmertalk] CW Skimmer - more

Wed Nov 12 17:06:58 EST 2008

It may be within the program.  As decoders come and go the memory they use
may need to be recycled, that would depend on how it was coded, what
language, and how the internal memory management interacts with windows.
That should be visible on one of the memory statistics in the task manager.

David Robbins K1TTT
e-mail: mailto:k1ttt at arrl.net
web: http://www.k1ttt.net
AR-Cluster node: 145.69MHz or telnet://dxc.k1ttt.net

> -----Original Message-----
> From: Pete Smith [mailto:n4zr at contesting.com]
> Sent: Wednesday, November 12, 2008 11:08
> To: k1ttt at arrl.net; skimmertalk at contesting.com
> Subject: RE: [Skimmertalk] CW Skimmer - more
> 
> I really don't think so, Dave.  Nothing else changes, and it does not
> appear that this issue is triggered by external events.  Instead, it is as
> if Skimmer itself has 2 different states - one in which it uses much more
> CPU than the other.  It toggles into the "gluttonous" state. usually at
> startup or when you stop and restart the waterfall or change the Sample
> Rate.  It will go from a lower CPU utilization to one that is much higher,
> while nothing else changes, either on the band being monitored or in any
> of
> the other programs.  Once it has switched, it stays high until some right
> combination of events causes it to toggle back into the normal CPU
> utilization.
> 
> 73, Pete
> 
> K1TTT wrote:
> >If what I suspect is right then it would only happen when windows needs
> >extra overhead time to swap processes.  Watch some of the other columns
> when
> >that happens, see if the pagefault or working set deltas are high when
> that
> >happens.  Those are signs that windows is having to do extra work to
> manage
> >memory, especially pagefaults.
> >
> >
> >David Robbins K1TTT
> >e-mail: mailto:k1ttt at arrl.net
> >web: http://www.k1ttt.net
> >AR-Cluster node: 145.69MHz or telnet://dxc.k1ttt.net
> >
> >
> > > -----Original Message-----
> > > From: skimmertalk-bounces at contesting.com [mailto:skimmertalk-
> > > bounces at contesting.com] On Behalf Of Pete Smith
> > > Sent: Tuesday, November 11, 2008 23:04
> > > To: skimmertalk at contesting.com
> > > Subject: Re: [Skimmertalk] CW Skimmer - more
> > >
> > > I don't have any trouble with the Task Manager number being larger,
> but
> > > why
> > > only some of the time?  Something makes the relationship between the
> two
> > > numbers flip unexpectedly, and when it does, my other programs slow
> down.
> > >
> > > 73, Pete
> > >
> > > At 05:53 PM 11/11/2008, David Robbins K1TTT wrote:
> > > >That could be extra windows overhead.  It is likely that the cpu time
> > > used
> > > >as sampled by the application can only report the actual time spent
> in
> > > it's
> > > >own process.  There are times however, especially if you are using
> the
> > > >computer for multi-tasking, when windows has to move things in and
> out of
> > > >memory.  This would be done outside of the actual program cpu usage
> so
> > > the
> > > >program built in meter may not register it, but the task manager
> would.
> > > You
> > > >could probably provoke such behavior by using another program that
> uses
> > > lots
> > > >of memory.  The less memory the skimmer has to work with the more of
> this
> > > >swapping in and out of memory is needed.  Look up terms like 'page
> > > faults'
> > > >or 'context swapping' and how windows does multi-tasking and memory
> > > >management.
> > > >
> > > >
> > > >David Robbins K1TTT
> > > >e-mail: mailto:k1ttt at arrl.net
> > > >web: http://www.k1ttt.net
> > > >AR-Cluster node: 145.69MHz or telnet://dxc.k1ttt.net
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Pete Smith [mailto:n4zr at contesting.com]
> > > > > Sent: Tuesday, November 11, 2008 22:34
> > > > > To: skimmertalk at contesting.com; SDR-IQ at yahoogroups.com;
> > > > > dxatlas_group at yahoogroups.com
> > > > > Subject: [Skimmertalk] CW Skimmer - more
> > > > >
> > > > > Now that I've proved (to myself at least) that I'm not completely
> > > nuts,
> > > > > let
> > > > > me prevail on your patience with a slightly more complicated
> question.
> > > > >
> > > > > Under normal circumstances, I understand that CW Skimmer's own CPU
> > > usage
> > > > > indicator will normally show a somewhat higher number than Task
> > > Manager's,
> > > > > because it is looking at peaks from one application while Task
> Manager
> > > > > looks at an average over a second of so of all the processes that
> are
> > > > > going on.
> > > > >
> > > > > That being said, I find that there is a pretty logical and
> explicable
> > > > > relationship among the Skimmer CPU number, the bandwidth set on
> the
> > > Radio
> > > > > tab, the number of decoders, and the number of candidate data
> streams
> > > (the
> > > > > last two are the "98 of 456 decoders" numbers).
> > > > >
> > > > > Incidentally, it's little-known, but the Sample Rate choice on the
> > > Radio
> > > > > tab *does* affect both the bandwidth covered by an SDR-IQ when
> used
> > > with
> > > > > Skimmer and (therefore) the CPU demand.  To verify this, bring up
> > > Skimmer
> > > > > and change from 48 to 96 KHz.  You'll see the CPU loading go 'way
> up.
> > > > >
> > > > > Finally, here's the question.  Have you ever noticed a situation
> in
> > > which
> > > > > the CPU utilization of the total of the two Skimmer processes) as
> > > reported
> > > > > by Task Manager is much higher than the CPU utilization being
> reported
> > > by
> > > > > Skimmer?  I see this all the time, but have not been able to
> figure
> > > out
> > > > > what causes it, or what makes it go away.  All I know for sure is
> that
> > > > > sometimes something is happening to impose a much higher load on
> the
> > > CPU
> > > > > for a given number of decoders than is the case at other times.
> > > Sometimes
> > > > > I'll see this as much higher CPU utilization in 48 KHz mode than
> in 96
> > > KHz
> > > > > mode.  Sometimes I can cause things to revert to normal by
> changing
> > > Sample
> > > > > rates, stopping and restarting the waterfall, or restarting the
> > > program.
> > > > >
> > > > > One good question that Alex asked was whether it might be a
> mistake by
> > > the
> > > > > Task Manager.  All I can say is that I don't think so, because
> there
> > > is a
> > > > > definite slowing of other programs when Task Manager says
> utilization
> > > is
> > > > > over 85 percent, even though Skimmer itself says it is only using
> 50-
> > > 55
> > > > > percent.  Moreover, I see the same phenomenon in Process Explorer,
> > > which
> > > > > is
> > > > > a completely separate program - I can't be sure that it doesn't
> > > measure
> > > > > CPU
> > > > > utilization the same way, of course, but it seems to be a stretch,
> > > because
> > > > > Process Explorer was not developed by Microsoft.
> > > > >
> > > > > Anyway that's the question - if you could, just keep your eyes
> peeled,
> > > or
> > > > > run some tests, whatever you have time for.
> > > > >
> > > > > Thanks and
> > > > >
> > > > > 73, Pete N4ZR
> > > > > the World Contest Station Database is back...
> > > > > www.conteststations.com
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > Skimmertalk mailing list
> > > > > Skimmertalk at contesting.com
> > > > > http://dayton.contesting.com/mailman/listinfo/skimmertalk
> > >
> > > _______________________________________________
> > > Skimmertalk mailing list
> > > Skimmertalk at contesting.com
> > > http://dayton.contesting.com/mailman/listinfo/skimmertalk
> 