Refining Cell Placement at Routing Stage

 refinePlace didn't do anything because your placement was already legal. 

Can you detect the routing congestion earlier in the flow, so you're not going back so many steps? Or does the routing congestion appear only after CTS, or only after postCTSOpt, or only after hold fixing? If so, the original placement may not be the problem, or at least not the whole problem. Just throwing out some ideas.

I can usually see where congestion spots will be, but I don't know which ones are going to be the worst routing bottleneck until after CTS and hold fixing. I usually just have one or two very small areas that end up being heavily congested to the point where it has trouble fixing all the routing violations. But 10 rows away, I might have any area where the placement utilization and routing congestion is very low. It would be nice to able to just automatically shift some of the cells from the congested area into the less utilized area. I currently have done this by hand and it has worked, but it's a pain because I have to do it manually.

Kari - you seem to be the expert so if you don't know of a way to do this I guess there propbably isn't one. :)

 if you know it is always the same module with congestion problem, you can use "specifyInstPad inst_name  integer".  You can use wildcards for the inst. name and you need to play around with the "integer" to get a good placement.

li siang

 There's plenty I don't know, believe me! :-) HOWEVER, I came across something: after you find your congested area, add a density screen (partial placement blockage) and then try placeDesign -incremental. Let me know if this works for you!

That's a good suggestion Li Siang. It sounds like thay may help with my problem. I will try that on my new project coming up. Thanks for the suggestion!

Kari said:

 There's plenty I don't know, believe me! :-) HOWEVER, I came across something: after you find your congested area, add a density screen (partial placement blockage) and then try placeDesign -incremental. Let me know if this works for you!

 

See - you DO know everything! :) This is what I was looking for. I think between this and trying what Li Siang suggested, I should be able to alleviated some of the congestion problems. Thanks!

 Hi Kari,

I just found this 2008 subject as I was searching informations on that congestion problem.

You wrote that there are severale solutions to solve congestion problems at cells placement. Could you please explain some others of these ?

 Thanks.

Phil

 Hi Phil,

 I don't see where I said that in this thread, but you should look into density screens (partial placement blockage) and cell padding (specifyCellPad). These are two typical ways to make sure the initial placement isn't too congested.

- Kari

 Thank you Kari,

I used cell padding and solve my congestion problem.

 Phil

 Phil,

That's great news! Thanks for posting back that you were successful.

- Kari