extract connectivity using skill

Harish,

In IC614 there's a new function to save the result of the "Mark Net" command to another cellView - leHiSaveAllHighLightMarkNet(). Unfortunately it doesn't really support a procedural approach to being called - it's just a UI function. However, with a bit of work, it can be made to respond to a procedural approach.

In order to mark the net you want, if you know the coordinates of somewhere on the net, simply do:

preXY(list(x y))
leHiMarkNet()
cancelEnterFun()

And then to save the result to a cellView, you can use CCSsaveMarkNet() as defined below:

procedure(CCSsaveMarkNetCB(lib cell view)
  let((form)
    form=hiGetCurrentForm()
    form->libName->value=lib
    form->cellName->value=cell
    form->viewName->value=view
    ;--------------------------------------------------------------------
    ; Form done needs to be done via a timer, otherwise the form
    ; gets stuck
    ;--------------------------------------------------------------------
    hiRegTimer(sprintf(nil "hiFormDone(%s)" form->hiFormSym) 1)
  )
)

procedure(CCSsaveMarkNet(lib cell view @optional (wid hiGetCurrentWindow()) "tttw")
  let((ddId)
    ;--------------------------------------------------------------------
    ; Delete the previous cellView to stop being prompted for overwrite
    ;--------------------------------------------------------------------
    ddId=ddGetObj(lib cell view)
    when(ddId ddDeleteObj(ddId))
    ;--------------------------------------------------------------------
    ; Fill in the form, via a timer - so is filled in after the save
    ; form is launched
    ;--------------------------------------------------------------------
    hiRegTimer(sprintf(nil "CCSsaveMarkNetCB(%L %L %L)" lib cell view) 0)
    leHiSaveAllHighLightMarkNet()
  )
)

All this can be done with VLS "L" licenses (no XL needed). You do need to ensure that you don't have a via (in the validVias constraint) which connects down to your diffusion layer, because otherwise connections will be made under the gate of  transistors by mark net.

If you're using earlier versions, you'll need to write a shape tracer, which is quite a lot of work...

Regards,

Andrew.

 If DIVA works then ivExtract would work for this. It saves the extracted data in a view extract.  Try it.

BTW, I filed an enhancement CCR number 740427 for a proper API for Mark Net.

Regards,

Andrew.

Hi Andrew,

Thanks for the reply. Unfortunately, iam Using an IC613 so i haven't got the idea of the function leHiSaveAllHighLightMarknet() and its features.

i'll give a brief explanation of what iam trying to do

1) Iam selecting a net in my top level (ex: vdd) by using xx = setof(x cvid~>shapes x~>net~>name == "vdd")

2) now i need to traverse in to and out of the hierarchy to select the connectivity of vdd in my instances.

3) once the net is traced i could use a dbCopyFig() function to copy the traced net.

the problem is that i could not find any functions that could help me trace a net hierarchically.

This code could be very helpful for me to find the shorts quickly. can u please guide me how i can do this

thankyou

Regards

Harish Poshala 

Harish,

One thing I don't understand - if you are using the base layout editor (not XL), how has the design got connectivity information present? Normally it's only XL (and GXL) which would add connectivity into the database?

Also, it should be straightforward to move from IC613 to IC614 - it's just an update release (no license changes needed), so that might be your simplest option.

You can select the shapes in the lower level cells by following the instTerms. There's some relevant posts on this before. For example Read and follow nets.Try searching a bit in the forums. I recall answering something similar elsewhere recently - it may have been here, or maybe on comp.cad.cadence (can't quite remember).

Regards,

Andrew.

Thanks Andrew,

i have now got an XL version which made my life easier. But my question was is there a way to extract connectivity information in VLE like in some previous post i read about dbGetTrueOverlaps() could do the trick, but was not able to find a way how this could do. if possible for u can u give me a hint of how actually it could be done.

I know the code would include a lot of work, but can u help me in finding a way to do it.  

thanks

Regards

Harish 

Check out this post if you haven't already.

/forums/p/11109/13204.aspx#13204

Derek

 Hi Harish,

I will add to what Derek and Andrew have said by providing an outline of how I would approach this in SKILL code:

1. Get a starting point (xy provided perhaps by enterPoint() or
   enterPoints()) or starting shape from the user
2. Obtain the object at the starting point (use gePointToFig()
   or dbGetOverlaps() with a depth of 0 for this)
3. Check that the object is on a routing layer
4. Add this object to the list of connected shapes
5. Use dbGetTrueOverlaps() to find objects overlapping this
   shapes' bounding box (decide the hierarchy depth you wish to
   use)
6. Filter the objects to remove non-adjacent connectivity layers
   compared to the current object, also filter any shape that is
   already on the list of connected shapes
7. Add these objects to the list of connected shapes
8. Repeat from Step 5. (recursive?) to follow connected shapes

The above is just an outline for finding the "connected" shapes in a similar fashion to the Mark Net feature. You may wish to add highlights to show what is being processed. The decision of whether something is connected will depend on the layers used, you may wish to use information from the Technology File (IC5141) or the viaDefs (IC61x) similar to how Mark Net works.

I hope that this information is helpful to you.

Best regards,

Lawrence.

Step #5 needs to be expanded, or else you can get incorrect results. If you have several L shaped polygons or paths, the bounding box of those shapes could overlap shapes which are not connected. You actually have to break all shapes into rectangles and then do the overlap check on each rectangle. This part is not trivial for polygons, although, as discussed in the linked post, dbLayerTile could help with this.

Derek

Yes, thank you Derek - I did read through the linked post but I forgot to mention about fracturing any input polygons in my post above. The suggested algorithm is one from my head and not proven by any actual coding!  dbGetOverlaps can indeed produce overlaps where the bounding boxes of shapes overlap but the actual shapes do not - dbGetTrueOverlaps might help but it might also be too much of an overhead (it will be slower since it is a less simplistic function).

Harish, do pay attention to Derek's comments since he has written functional code on this topic and I have not.

Best regards,

Lawrence.