Offering help with the DFT tool

[MohamedWHassan]

Greetings,

I'm a researcher at Tampere University, and I wanted to further develop the DFT tool.
I wanted to get your insight on what the tool might be lacking so far and if there are other people working on it so that our work wouldn't overlap.

As far as I understand the tool is currently lacking scan chain reordering and compression/decompression, could I contribute to either of these?

Thanks,
Mohamed

[maliberty]

Contributions are definitely welcome. Prof. Kahng at UCSD wrote up some gaps in this doc. There is an earlier spec here (not certain about permissions, let me know if it doesn't open).

Both of those areas are open. It would be good to do a small write up of your plan to make sure we are in sync and don't wind up with surprises.

@fgaray is the primary author and can give guidance as well.

[fgaray]

Could you please request it from docs's UI? Thanks!

[MohamedWHassan]

Requested

[MohamedWHassan]

Thanks for sharing the file @fgaray.
I have been through the DFT specification documents, the source code for the DFT tool and this course about OpenDB and I had a couple of questions.

1- I was wondering if you would recommend some DFT gap that I can work on which wouldn't be too difficult as it is my first time contributing to large code base projects.
2- Are there any other resources that you use to when interfacing with OpenDB because I don't see a documentation for it or do you just go through the source code?
3- Is it best to communicate here, or do you have a preferred channel?

[maliberty]

db.h describes the public interface for opendb. I'm glad to answer questions you may have about it here.

@fgaray do you have a suggestion for (1)?

[fgaray]

Hi Mohamed,

I don't know if you also have access to this doc but we discussed that when architecting scan chains, you should specify a set of scan chains to be created (see the solved comments there). This config should be the hard number of chains to create.

Currently we have a setting similar to that called "-max_chains". We need to:

1. Verify what is really doing in code. The README may be missleading.
2. Rename it to "chain_count" (or something similar to that) (Maybe not remove -max_chains as there may be users of that so maybe let's warn them for a while? @maliberty how do we do releases or breaking changes like this one?)
3. Modify the implementation so scan architect creates the requested number of scan chains. If there are scan chains without flops, maybe print a warning.

Re 3) This is fine or you can write me an email (you can find it in git).

Thanks for working on DFT!

[maliberty]

There is some work from UCSD that could be integrated at https://github.com/ABKGroup/OpenROAD/tree/dft. I haven't had a chance to look at it more closely but if you are interested let me know.