• Previous message: R. Timothy Edwards: "Extract error"
• In reply to: R. Timothy Edwards: "Extract error"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Tim,

Thanks for the feedback and work around on this.

My fix was simply to break the compound device into multiple, independent devices
without common source/drain diffusions.  I did keep the poly gates as one piece of
poly.  Fortunately I had plenty of room to make the changes.

Also, I agree that this is a rare condition.  I've been using magic several years now
for extractions on designs with 1K to 40K transistors and this is the first time it's
had a problem that wasn't in the cockpit.

Best regards,

Cecil

"R. Timothy Edwards" wrote:

> Dear Cecil,
>
>    I haven't really got to the bottom of the magic extract error you
> described, but at least I can flag a warning, ignore the situation
> and continue on with the extract, and it seems to work okay and
> produce a valid extraction.
>    I am updating the magic source database, but you can make the
> fix yourself as follows:
>
> subdirectory extract, file ExtBasic.c, routine ExtSortTerminals(), line 528, add:
>
>                 else if( p2->pt.p_x == p1->pt.p_x && p2->pt.p_y == p1->pt.p_y )
>                 {
>                     TxPrintf("Extract error:  Duplicate tile position, ignoring\n");
>                     continue;
>                 }
>
>    The magic extractor apparently does not expect to see two tile positions
> with the same coordinates, which suggests that it was in error when it
> created the structures.  However, as it appears to be an extraordinarily rare
> case, I will just treat the symptom and not the disease.  It might, however,
> be instructive to know how you changed the layout that prevented the extract
> from hanging, as you mentioned.
>
>    This error occurs on your layout for all versions of magic, as far as I
> can tell, including the current development version, 7.1.
>
>                                                         Regards,
>                                                         Tim
>
> >    I don't know if you are interested in bug reports on 6.5.1 but here
> > is a weird one I ran across.
> >
> >    I have a geometry that causes Magic to hang up in the process of extracting
> > a circuit from the layout.  It consists of a series of nmos devices
> > with a common gate. The top three devices are shorted together.
> >
> >    If I use the extract command magic goes off into never-never land, chasing
> > around in a loop somewhere.  However, if I rotate the geometry by
> > 90, 180 or 270 it does what it is supposed to do.
> >
> >    Ever heard of anything like this?
> >
> >    Regards,
> >    Cecil Aswell

• Next message: Steve Tell: "subscribe"
• Previous message: R. Timothy Edwards: "Extract error"
• In reply to: R. Timothy Edwards: "Extract error"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]