Trying to compare 2 GDS files using dbdiff command

Hi,

Just to check the dbdiff command, i streamed out the same cadence OA database twice at 2 different directories. Tried to compare those 2 GDS files, it is saying the cells are different.

It looks strange to me. Could anyone help me on this, if you have faced similar issue? what could be wrong in this scenario?

Do let me know, if the data is insufficient.

Thanks,
Arunkumar A

Arunkumar A said:

Hi,

Just to check the dbdiff command, i streamed out the same cadence OA database twice at 2 different directories. Tried to compare those 2 GDS files, it is saying the cells are different.

It looks strange to me. Could anyone help me on this, if you have faced similar issue? what could be wrong in this scenario?

Do let me know, if the data is insufficient.

Thanks,
Arunkumar A

Click to expand...

This is not surprising (as it may seem from the first glance) - GDS file format contains the date/time of last modification, as well as date/time of last access - with a time resolution down to seconds - in BGNLIB record, as well as creation time and last modification time in BGNSTR records.

Thus, even your layouts are identical - GDS files will be different, because of the difference in stream-out time.

GDS file format is explained, for example, in this document:
http://bitsavers.informatik.uni-stuttgart.de/pdf/calma/GDS_II_Stream_Format_Manual_6.0_Feb87.pdf

Max
----------

timof said:

This is not surprising (as it may seem from the first glance) - GDS file format contains the date/time of last modification, as well as date/time of last access - with a time resolution down to seconds - in BGNLIB record, as well as creation time and last modification time in BGNSTR records.

Thus, even your layouts are identical - GDS files will be different, because of the difference in stream-out time.

GDS file format is explained, for example, in this document:
http://bitsavers.informatik.uni-stuttgart.de/pdf/calma/GDS_II_Stream_Format_Manual_6.0_Feb87.pdf

Max
----------

Click to expand...

Hi Max,

Thanks for the info. I have 2 questions.

1) You said the time stamp would be different but the database is same right? The GDDS compare tool [dbdiff] should check the layer information rather than the time stamp. Is my understanding right?
2) Can you suggest me a tool which can do better comparison in this scenario?

Thanks in advance

Hi -

Sorry, I am not familiar with dbdiff tool - I guess this is a tool from Mentor? Can you contact them for technical support?

If dbdiff ignores the dates, and checks only the content of the layout - then I agree with you that this situation is strange...

I would suggest you to check this tool: klayout (www.klayout.de), Tools-> Verification-> Diff tool (or XOR tool)

Max
---------

timof said:

Hi -

Sorry, I am not familiar with dbdiff tool - I guess this is a tool from Mentor? Can you contact them for technical support?

If dbdiff ignores the dates, and checks only the content of the layout - then I agree with you that this situation is strange...

I would suggest you to check this tool: klayout (www.klayout.de), Tools-> Verification-> Diff tool (or XOR tool)

Max
---------

Click to expand...

Hey Hi,

I got to know that simply using dbdiff is not the best practice. It is always better to compare the layers. So using dbdiff command we can get xor rule file and then we can run calibre to find the difference. Below are the commands.

dbdiff -system GDS -design <GDSNAME TOPCELLNAME> -refdesign <GDSNAME TOPCELLNAME> -write_xor_rules <XOR_RULEFILENAME>
calibre -drc -hier -turbo -hyper -64 <XOR_RULEFILENAME>.

Arun

Hi Arun -

thanks for the info - interesting!

This makes a lot of sense to me - writing a special rule file where you define how the two GDS files should be compared.
Without that - how would a software tool to know how the geometries should be compared?
The content of GDS files may be different - such as PATH vs BOX, but the geometry may be the same.

The upside of this approach (dbdiff) is the power of Calibre.

The downside - you need to know SVRF language, and you need to have Calibre software and licenses.

The approach I suggested - using klayout functionality - may not be as powerful, but it is quite powerful, and it's free and a pleasure to use.

Maxim
-------------