How to connect endcap cell vertically in soc encounter?

Hi Renee,

 You're a busy designer! That's great and I'm glad you've joined the forum to ask your questions. :-)

I don't think your picture got attached, but from your description, I think I can say that SRoute is not meant to do what you are attempting here. It is meant to connect the horizontal rails as it did, but the vertical connection you mention doesn't sound like a typical flow, and a picture would definitely help me see what you are trying to do.

 - Kari

I am sorry that although I inserted the file, it seemed that it is not visible, and I don't know why.

My idea is :

 It is like after adding stripes, there are power stripes going through the whole design vertically, I want my design to have endcap cells(ENDCAPL(L for left, as Vdd! connection) and ENDCAPR (R for right, as GND! connection)) connected vertically, then in the hierachical view, it looks as if there is a stripe to connect the Vdd! and GND! net from the top of the design to the bottom. 

I saw that some information about endcap cells the soc encounter user guide, it said that "They are placed at the ends of the site rows, and are used in some technology for power distribution." So, is it possible to force SRoute to route vertically for endcap cells?

Thank you for your help!

Renee

Renee, would you mind trying to insert the image again?  I want to be sure the system is functionaling correctly, or know if we have a problem.  To insert an image you use the options tab and attach a file.  Images should appear inline.  Other file types as linked attachments.  Here is an example of at attached inline image:

communityops 

I did upload the endcell.jpg, but when I press the tab post, it did not post the message.

Renee,

 I'm not sure what you want these vertical stripes to do that a regular vertical stripe doesn't do already. Is it just that you want them to line up with your columns of endcap cells? I think you could accomplish that with the regular addStripe command by fine-tuning the starting offset and spacing. 

- Kari 

I finally can upload the picture.

As shown, the clock signal go vertically into leftend cell, but is distributed horizontally through the overall row, and leftend cells on end of the row will connect to other leftend cells on neighbour rows. For those combinational cells, the clock signal and scan signal will only go straight through them without connection, and those signals will connect by abutment.

But I found that it seemed that soc encounter did not support cells with horizontal pins in the middle of the cell. I added clock and scan signals to the combinational cells, and at the place cells stage some cells cannnot be placed and some of those placed cells are overlapped.

I don't know why.

 Hi Renee,

I'm glad the picture uploading is finally working for you! This design looks different than what I'm used to dealing with, but I would think that the small connections (between a left endcap and the one below it) should be handled just fine by nanoRoute. If I understand correctly, you would like the thicker (in the picture, maybe they aren't really thicker than min width) vertical clock routes to be done by SRoute or another automated way. If there are not a lot of these, it may be pretty easy to draw them by hand. Otherwise, I still think you can use addStripe to route them, but it takes some calculating to get the right offsets and spacings.That still leaves the horizontal taps into the vertical clock bus. I would probably let nanoroute attempt these as well. Does it sound like any of this will work for you? 

- Kari 

 I forgot to address your placement issues. Can you post a picture of the overlap? I'm not really sure what it is about the cells that would cause issues during placement. 

- Kari 

The nanoRoute did not route the small connection between a left endcap and the one below it. Is it because that I did not provide wide wire to it? As there are not many of these connection, I might connect them by hand.

About the placement problem, now all the cells can be placed, but although I tried to add filler cells to fill the gap, there is still some gap between cells, and I believed that it is not the problem of the filler cell's width( I did add filler cell before, and there is no problem with it). In my own cell library, the clock, nReset and scan signal go through them horizontally, and for the combinational cells, these signals go straight through without connection, and I want the clock, nReset and scan signal connected by abutment. But, during the placement, it seemed that there was space between each cell, and it cannot be filled by the filler cell.Moreover, there were drc violations for clock, nReset and scan signal's routing.

Hi Renee, 

 In general, if you have nets that are connected ONLY by abutment, then you should put a -skip_routing attribute on them:

setAttribute -net net1 -skip_routing true

It's hard to tell from just a picture, but it seems like nano is trying to route your abutment nets, and in doing so, crosses the cells, creating violations. Does that seem like what's going on to you? Do your filler cells continue the clock, nReset, and scan connections? Do you add them before routing? If fillers aren't being added to the gap, and there's nothing blocking them (like routing or placement blockage), then it's usually because you don't have a filler cell that's small enough. But you said they got put in correctly before, so something else may be going on.

For the other routes (between an endcap and the one below it), was your hand-routing successful? Sometimes trying to route by hand will show the cause of why the router couldn't connect the net. 

- Kari