Select layer connected to other layer

kareemfarid

Suppose that I have a set of connect statements, for example:

    connect(m1, via)
    connect(via, m2)
    connect(m2, via2)
    connect(via2, m3)
    connect(m3, via3)
    connect(via3, m4)
    connect(m4, via4)
    connect(via4, m5)

How to select m1 shapes that are connected to m5?

Matthias

Hi @kareemfarid,

If you are only interested in m1 shapes being somehow connected to m5, the easiest way is not to use "connect", but to traverse interactions downwards from m5:

via4_connected_to_m5 = via4.interacting(m5)
m4_connected_to_m5 = m4.interacting(via4_connected_to_m5)
via3_connected_to_m5 = via3.interacting(m4_connected_to_m5)
m3_connected_to_m5 = m3.interacting(via3_connected_to_m5)
...
m1_connected_to_m5 = m1.interacting(via_connected_to_m5)

Matthias

kareemfarid

Thanks @Matthias. Is there any other way to do so?. Assuming a larger connectivity stack, I would have to use a lot of interacting statements. I thought about using a separation with props_eq with a very large value, but probably this will have performance penalty.

Matthias

The number of interacting statements should the same than the "connect" statements. Why do you think you need a larger number?

Doing a separation with a large value isn't a good idea. That will definitely be very slow.

But I understood your requirement to identify all M1 that are in some way connected to M5. From your proposal I guess you actually want something else. Are you trying to build an antenna check?

Matthias

kareemfarid

The test case above is a small example. I am trying to implement multiple checks that depends on same/different net. In addition to these, there are checks where one layer has to be connected to another. In the first case connect is particularly useful. I was wondering if connect will be useful in the second case too. I guess I can duplicate the connect statements with similar interacting statement.

Matthias

Hi @kareemfarid,

maybe some example is helpful to demonstrate what you intend to do. The initial post was going in a different direction.

For same/different net separation, you need "connect" as only that enabled net annotation.

The usual application are DRC checks between same or different nets and antenna checks.

It is essentially possible to extract individual nets and analyze them on a per-polygon level. However, that will be pretty slow. Internally, the antenna checks do exactly that when they compute antenna ratios, but their loop executes in C++ much more efficiently.

Matthias

kareemfarid

@Matthias I am implementing a couple of checks for sky130 PDK. Specifically:

• hv.poly.6a: Min spacing between nwell and deep nwell on separate nets
• hv.diff.1b: Minimum spacing of (n+/p+ diff resistors and diodes) connected to hv_source/drain to diff

and other similar rules that require "... connected to .." The connectivity can be established using the entire metal stack so I thought connect statements might be useful here, which I was going to use anyway for hv.poly.6a and other similar same/different net checks.

blueman_44

Like Matthias mentioned antenna_check computes the areas of the specified layers , which are connected, so if you set the ratio value to 0 then it should grab all shapes of the layer connected to those nets. Here's an example using antenna_check to find floating poly:

report("drc")

#input layers
poly = input(1,0)
contact = input(2,0)
metal1 = input(3,0)

#connectivity
connect(poly, contact)
connect(contact, metal1)

poly_metal1_conn = antenna_check(metal1, poly, 0.0)              #selects all poly shapes connected to metal1
(poly - poly_metal1_conn ).output("floating poly to metal1")      #outputs floating poly 

Matthias

Thanks @blueman_44, that is smart

@kareemfarid - I think we are talking net classification here. Apparently the goal of the rule is to identify high voltage nets and demand a bigger spacing for those.

I don't have a good solution for this yet. I guess there is a missing feature to implement this efficiently. A kind of net tagging comes to mind.

Matthias