How EL/RF Pin Capacitance values obtained from .lib?

[junseokhur]

Hello, thank you for sharing your wonderful work.

I'm having trouble finding out EL/RF output pin capacitance feature from .lib files you provided.

For input pin, I can find EL/RF capacitance from sky130_fd_sc_hd__ff_n40C_1v95.lib, sky130_fd_sc_hd__ss_100C_1v60.lib.
However, for output pin, there is only max_capacitance value provided in .lib files.

I'm wondering how did you obtain capacitance values of output pins.

[gzz2000]

Hi, thank you for your interest in our work!
Actually in NLDM cell delay modeling, output pins do not have defined capacitances. Only input pins have capacitances defined. As cell delay is modeled as a function of output load and input slew, the load of output pin itself are modeled as zero.
The max_capacitance is not the capacitance of output pins -- they are constraints on output load.

[junseokhur]

Thank you for your fast reply!!!!

[junseokhur]

I have another trouble.
I tried to read lef file (merged_unpadded.lef) in DREAMplace to find out pin_coordinate but LEF parsing failed with error below.
Is there any reason why DREAMplace cannot parse lef file used in this work?
Thank you.

[gzz2000]

Try removing the excessive slash / after .lef?

[junseokhur]

Yes I tried but same error appears

[gzz2000]

I would still guess this is a issue related to file path or file read permissions. Can you double check you can access the file with that path?

It may be helpful to set a breakpoint around the DREAMPlace read logic and see if any file read error is reported.

[junseokhur]

Thank you for your advice. I'll check how DREAMplace parse lef file.

Sorry for too many questions but I have one last question :(

In the case of PDKs other than sky130 (asap7), the LUT of the cell's output pin is divided in detail according to the input vector as well as the related pin. For example, when the input pins are A and B, different LUTs are defined for A&B and !A&B. In the case of this PDK, I would like to ask for advice on how to embed the LUT in the cell edge feature

[junseokhur]

Problem solved thanks

[young03600]

Hi,
I would like to know how the pin coordinates are extracted using dreamplace?

Thanks.

[gzz2000]

Hi, I would like to know how the pin coordinates are extracted using dreamplace?

Thanks.

Hi, you can read the DEF into DREAMPlace, and call the pinpos operator to get a tensor of pin positions. You can read DREAMPlace's HPWL operator to see how its inputs are constructed (their input contains the pin coordinate tensor).

[young03600]

Hi, you can read the DEF into DREAMPlace, and call the pinpos operator to get a tensor of pin positions. You can read DREAMPlace's HPWL operator to see how its inputs are constructed (their input contains the pin coordinate tensor).

Hi,
Many thanks for your reply.

I've retrieved the pin locations but locations do not matched up with the dumped nf[2:5] (I used the desgin "spm" for test)
But I got slightly different pin locations.

BTW, I checked the description of nf[2:5] in the repo.
But I found the order seems not to be top/left/right/bottom of die area.
The nf[2]+nf[5] and nf[3]+nf[4] not equal to the height and width of die area respectively.

Thanks.