Slow ocnPrint

[Adam]

I'm having problems with a really slow ocnPrint. I am running a quite
long transient simulation. The simulation itself takes only a few
minutes but the printout to a text file using ocnPrint takes several
hours (the resulting file is in the order of 10 MB). So most of the
time is spent waiting for ocnPrint to do the printing.

Is there any faster way to export the simulated data into a text file
or is there a way to speed ocnPrint up?

Many thanks for any help on this.

/Adam

 

[Svenn Are Bjerkem]

In article <1132850173.082039.174890@g49g2000cwa.googlegroups.com>,
adam_strak@yahoo.com says...

I'm having problems with a really slow ocnPrint. I am running a quite
long transient simulation. The simulation itself takes only a few
minutes but the printout to a text file using ocnPrint takes several
hours (the resulting file is in the order of 10 MB). So most of the
time is spent waiting for ocnPrint to do the printing.

Is there any faster way to export the simulated data into a text file
or is there a way to speed ocnPrint up?

In this case I would recomend to try the "psf" binary.
$> psf -h
give you about the only documentation available for this utility, but
when you know the name of the signal(s) you want to investigate, then
you can dump these to a file as ascii or as a new psf.

The ascii dump from psf needs a bit "massaging" in order to be as
readable as the output from ocnPrint. I would guess that it is possible
to use awd or wavescan in stand alone mode on that extracted psf binary
to use the print or table function to get it right in case you don't
want to use awk/perl/tcl/ruby/python (or whatever) to do the job.

I would guess that the job is a bit larger in case you have a parametric
or corner simulation as the results are stored in separate directories
and scoreboardes in the runObjFile. (I would use tcl to parse the
central runObjFile to extract the loaction of the individual result
files and then fire psf on each one of them. Problem is just to join
these separate files to create the family.)

A third possibility would be to check out some of the psf related
functions in skartistref.pdf to see if it is possible to roll your own
ocnPrint
--
Svenn

 

[Andrew Beckett]

On 24 Nov 2005 08:36:14 -0800, "Adam" <adam_strak@yahoo.com> wrote:

I'm having problems with a really slow ocnPrint. I am running a quite
long transient simulation. The simulation itself takes only a few
minutes but the printout to a text file using ocnPrint takes several
hours (the resulting file is in the order of 10 MB). So most of the
time is spent waiting for ocnPrint to do the printing.

Is there any faster way to export the simulated data into a text file
or is there a way to speed ocnPrint up?

Many thanks for any help on this.

/Adam

If you use any of the ?from ?to or?step arguments, it will be slow, because it
is having to do clipping and interpolation. If you output the data directly it
should be pretty quick...

How are you using ocnPrint?

Andrew.

 

[Adam]

Hi!

I am not using any ?from or ?step. An example I have been working with
involves an inverter's transient simulation. The printout with ocnPrint
is according to:

ocnPrint(VT("/INV1_in") VT("/INV1_out") ?output "./cadence_data.txt"
?numberNotation 'engineering ?numSpaces 10 ?precision 10)

Actually, one printout took at least 10 hours whereas the simulation
took something like 10 min. Can the problem be that I am formatting the
output with ?numberNotation or ?numSpaces etc? Right now, I am looking
into psf as Svenn suggested but the output I am getting using e.g. "psf
-s -i timeSweep.tran -o INV_in.txt -t INV1_in -t t -f "%d %d" " isn't
easy to interpret. Maybe I am using the psf command wrongly? (the psf
help is a bit brief)

Thanks!

/Adam

 

[Svenn Are Bjerkem]

In article <1133190673.339536.132910@z14g2000cwz.googlegroups.com>,
adam_strak@yahoo.com says...

psf -s -i timeSweep.tran -o INV_in.txt -t INV1_in -t t -f "%d %d"

skip the -f at the end. %d translate the floating point number to an
integer.

Have you checked if you can use print or printvs from the calculator?
--
Svenn