How to ignore some minor dots in streaming out the gds file

Which version of the tools are you using?

Regards,

Andrew

Thank you, Andrew.

1. I used  /IC5.10.41.500.3.34.lnx86 to stream out two gds files

2. Then used dbdiff -system GDS -design xxx.gds xxxl -refdesign yyy.gds yyy -rdb dbdiff_result.rdb to generate rdb file

     ( /cad/mentor/2011.1/ixl_cal_2011.1_15.11/bin/dbdiff )

3. Then in Virtuoso open Calibre RVE to see the result

4. As I mentioned I found thounds of tiny dots in the layout that cause the difference between two gds files

Thank you !!

OK, so the "dots" are presumably small differences between the two layouts. I'd expect with RVE you can view the two GDS source files in the regions of where these dots are so you can investigate how they are different? I'm guessing that there is some small differences between your two layouts and this might help you find where they are.

I don't think "removing small dots" will help you here if the small dots are coming from differences between the two (e.g. small rounding error differences in pcells - who knows/?).

Put another way, I'm not sure how anyone could advise you without seeing the two GDS files (and preferrably the source layouts).

The IC version you are using is really old too...

Andrew.

So the dots are the output of a diff tool? They may not be a part of the database but rather a difference in rounding floating point numbers when creating the GDS. Check where the dots are and if they are connected to any shapes in the database.

Ted

Actually between the original two gds files I ca not tell anything different (by my bare eyes) in the place where the RVE shows the difference (the dots). For instance, one of errors locates inside a big area of metal 5; there should be no difference in metal 5 there between the two layouts but the result does show a error there. 

So maybe it is due to what you mentioned a difference in rounding? 

If yes it is, how to avoid creating such an rounding?

Thank you ! 

It might be rounding in dbdiff (a Mentor tool) for all we know - as I said, hard to know what to suggest if you don't know the  root cause of the difference...

Andrew.

Yeah, here is not a forum for Mentor tool, but do you know how to prevent the rounding in dbdiff?  :>

Or is there another way in Cadence flow to do the LVL without this "rounding" error?

Many thanks !! 

You could do the comparison with PVS (the Cadence Physical Verification System). Of course this is assuming that the difference is due to the comparison - I'm guessing wildly here in the absence of seeing the data for real.

Andrew.