DFT: Expected def syntax for scanchain startpoint/endpoint when connected to a component ?

[Essenceia]

Hello,
I am currently working on adding a scan chain to a design connected to a JTAG TAP, as such, unlike what seems to be the most common usecase, the scan chains startpoint, enable, and endpoints are not a top level pin but connected to an internal component.

I am opening this as a discussion as I am not certain if the bug is between the keyboard and the chair, in the def file or in the odb def file reader.

In the current behavior the scan chain start/endpoint connected to this internal component are written as :

+ START PIN m_ff_sc_tdi_buf/X
+ STOP PIN m_ff_sc_tdo_buf/A ;

Based on my understanding of the scan chain def syntax according to LEF/DEF 5.8 Language Reference, this format seems to treat them as I/O pins ?

START {fixedInComp | PIN} [outPin]
Specifies the start point of the scan chain. You must specify this point. The starting point
can be either a component, fixedInComp, or an I/O pin, PIN. If you do not specify
outPin, the router uses the pin specified for common scan pins.

STOP {fixedOutComp | PIN} [inPin]
Specifies the endpoint of the scan chain. You must specify this point. The stop point can
be either a component, fixedOutComp, or an I/O pin, PIN. If you do not specify
inPin, the router uses the pin specified for common scan pins

I am wondering if the expected format should not rather be ?

+ START m_ff_sc_tdi_buf X
+ STOP m_ff_sc_tdo_buf A ;

Note: I am not asking for a fix.

Best

openroad version: dcf36133a369abc8f3c5e5738cd4d82e4903c0e0

• gui: minor cleanups in Chiplet3DWidget #9486

[gadfort]

The format you are suggesting is correct per the spec:

We probably need to fix that otherwise other tools will not be able to parse the DEF from OpenROAD.

[Essenceia]

Perfect thanks. 👍

[maliberty]

This looks like a bug; I'll look at fixing it.

[Essenceia]

If you haven't already started fixing it can you leave it to me ? I fell I owe the project one.

[maliberty]

Thanks for the offer but I already did #9706. I'm sure there will be more chances to contribute 😄

[Essenceia]

Haha, can't speed past the main contributor on his own codebase, next time I am just opening the PR. 🤣
Thanks for your work, I will go give the fix a spin.