Wavescan keep on taking the focus

S

Svenn Bjerkem

Guest
Hi,

I have a fairly long skill script that evaluates a lot of simulations
that I have done. Nothing special about that. The special thing (or
very typical ting for skill) is that it is slooooow. This means that it
take a long time to generate all plots. That would also not have been a
big problem if that wavescan accepted that I wanted it to stay in the
_background_ and don't bother me until it is finished.

But wavescan keeps on taking the focus for every waveform it wants to
plot. It just won't go away. Is there a way to work around this?

--
Svenn
 
On 20 Sep 2006 07:46:12 -0700, "Svenn Bjerkem" <svenn.are@bjerkem.de> wrote:

Hi,

I have a fairly long skill script that evaluates a lot of simulations
that I have done. Nothing special about that. The special thing (or
very typical ting for skill) is that it is slooooow. This means that it
take a long time to generate all plots. That would also not have been a
big problem if that wavescan accepted that I wanted it to stay in the
_background_ and don't bother me until it is finished.

But wavescan keeps on taking the focus for every waveform it wants to
plot. It just won't go away. Is there a way to work around this?
I recall this being a bug with the Java version used. Which IC subversion are
you using?

I think this was a problem around IC5141 USR3 timeframe if my memory is correct.
More recent ISRs don't seem to have this problem - I think it was when the Java
version was changed.

And SKILL is not "typically" slow - I've written big applications, processing
a lot of data, and it can be very fast. It just depends on what you're doing
and how you're doing it... Obviously it's not as fast as a compiled language,
but it's still a virtual-machine based language, and is fairly fast - unless you
end up generating a lot of garbage (I mean that in the programming sense,
not as a term of criticism!).

Regards,

Andrew.
--
Andrew Beckett
Principal European Technology Leader
Cadence Design Systems, UK.
 
Andrew Beckett wrote:
On 20 Sep 2006 07:46:12 -0700, "Svenn Bjerkem" <svenn.are@bjerkem.de> wrote:
[snip]

But wavescan keeps on taking the focus for every waveform it wants to
plot. It just won't go away. Is there a way to work around this?

I recall this being a bug with the Java version used. Which IC subversion are
you using?
WaveScan version IC6.1.0EA.162 stand alone on the command line with
data generated with spectremdl spectre (ver. 6.1.0.151 -- 26 May 2006).

And SKILL is not "typically" slow - I've written big applications, processing
a lot of data, and it can be very fast. It just depends on what you're doing
and how you're doing it... Obviously it's not as fast as a compiled language,
but it's still a virtual-machine based language, and is fairly fast - unless you
end up generating a lot of garbage (I mean that in the programming sense,
not as a term of criticism!).
Ok, SKILL is not that slow, and it is faster to smash up a script to do
some manipulation than to write out the data to file and do the same in
excel. I was probably more thinking of how long time it take to
calculate and plot waveforms in the java version of wavescan (I guess
there is only one version of wavescan and that is the Java one, or do
we have some goodies coming up on our radar screens?)

This could also be a result of the way that I use wavescan with command
line spectremdl. If wavescan has to reinterpret all the logFiles in the
..raw directories every time it calculates a new waveform, then I can
understan why it take so long time. Then it would probably help to
reduce the number of saved signals, which I will go back to the
simulation and test out. So far I have been debugging my plots and have
occasionally needed one or the other signal.

--
Svenn
 
On 22 Sep 2006 01:48:33 -0700, "Svenn Bjerkem" <svenn.are@bjerkem.de> wrote:

Andrew Beckett wrote:
On 20 Sep 2006 07:46:12 -0700, "Svenn Bjerkem" <svenn.are@bjerkem.de> wrote:


[snip]

But wavescan keeps on taking the focus for every waveform it wants to
plot. It just won't go away. Is there a way to work around this?

I recall this being a bug with the Java version used. Which IC subversion are
you using?

WaveScan version IC6.1.0EA.162 stand alone on the command line with
data generated with spectremdl spectre (ver. 6.1.0.151 -- 26 May 2006).
Hmm, whilst that is not a released version of the tool - it's an "early adopter"
version - IC61 is not released until the end of October - I would have thought
that the Java version was recent enough.

And SKILL is not "typically" slow - I've written big applications, processing
a lot of data, and it can be very fast. It just depends on what you're doing
and how you're doing it... Obviously it's not as fast as a compiled language,
but it's still a virtual-machine based language, and is fairly fast - unless you
end up generating a lot of garbage (I mean that in the programming sense,
not as a term of criticism!).

Ok, SKILL is not that slow, and it is faster to smash up a script to do
some manipulation than to write out the data to file and do the same in
excel. I was probably more thinking of how long time it take to
calculate and plot waveforms in the java version of wavescan (I guess
there is only one version of wavescan and that is the Java one, or do
we have some goodies coming up on our radar screens?)
Wavescan is also quite fast - people often assume that Java is slow,
mainly from the historical problems there were with early Java versions.

Wavescan is being moved to Qt in phases - but that won't be until
next year.

This could also be a result of the way that I use wavescan with command
line spectremdl. If wavescan has to reinterpret all the logFiles in the
.raw directories every time it calculates a new waveform, then I can
understan why it take so long time. Then it would probably help to
reduce the number of saved signals, which I will go back to the
simulation and test out. So far I have been debugging my plots and have
occasionally needed one or the other signal.
If you're using wavescan standalone in MDL, then that's not using SKILL
at all. Note that it doesn't have to read the entire results directory to
plot results - it can selectively load and plot particular waveforms (this is
also true with SKILL mode wavescan).

If you're having performance problems, please contact customer support so we can
take a look.

Andrew.
--
Andrew Beckett
Principal European Technology Leader
Cadence Design Systems, UK.
 
Andrew Beckett wrote:

If you're using wavescan standalone in MDL, then that's not using SKILL
at all. Note that it doesn't have to read the entire results directory to
plot results - it can selectively load and plot particular waveforms (this is
also true with SKILL mode wavescan).
Hi Andrew,
I am starting wavescan -expr skill to access data which have been
generated by spectremdl run standalone and I now do most of my skill
development in such a wavescan session. Mostly due to the very much
improved behaviour of the CIW.

--
Svenn
 

Welcome to EDABoard.com

Sponsor

Back
Top