[Magic-dev] subcell CIF/GDS interactions
R. Timothy Edwards
tim at multigig.com
Wed Nov 30 14:01:48 EST 2005
I have just posted magic-7.3.114. This revision fixes an error that
has been a longtime problem in magic, and is actually a rather trivial
fix. This means that although I have applied the fix to 7.3, someone
(else, preferably) should generate a patch against version 7.1.
Previously, arrayed subcells that were too narrow and closely spaced
would generate a warning about possible bad CIF/GDS interaction between
non-neighboring cells with the potential to write bad CIF/GDS output.
This was particularly bad in that the resulting layers can't be seen
with the "cif see" command (although you can see them with the "cif *array"
command, except that the command option isn't widely advertised).
Usually this does not result in an actual output error, so I have been
used to ignoring the warning message. However, every now and then I
get real erroneous output. Tracking one of these instances down, I
realized that it happens when any cif layer in the tech file has a "grow"
distance larger than the cell width in the direction of the arrayed
instances. Interactions are then found between the 1st and 3rd cells
in the array (or 4th, or more). Magic optimizes the search for inter-
array interactions by finding interactions between the first two cells
and copying them enough times to cover the array. This assumes that
the interactions are only between the first two cells of the array.
If interaction paint has been generated relative to, say, the 3rd cell
in the array, this gets copied off the end of the array.
The solution is actually quite simple: reduce the size of the array
to at most 2x2 before computing interactions, and then restore the
original size of the array at the end.
The function cifHierArrayFunc() in cif/CIFhier.c is otherwise unchanged
from magic version 7.1 and can simply replace the original.
Regards,
Tim
+--------------------------------+-------------------------------------+
| Dr. R. Timothy Edwards (Tim) | email: tim at opencircuitdesign.com |
| Open Circuit Design, Inc. | web: http://opencircuitdesign.com |
| 22815 Timber Creek Lane | phone: (301) 528-9364 |
| Clarksburg, MD 20871-4001 | cell: (240) 401-0616 |
+--------------------------------+-------------------------------------+
More information about the magic-dev
mailing list