Description

In the earlier releases of the PDK (e.g. 3af133706e554a740cfe60f21e773d9eaa41838c), the naming conventions are kept the same with the commercial release. However, in newer releases the names are modified to be consistent with the previous SKY130 OpenPDK.

Investigation

In release 3af133706e554a740cfe60f21e773d9eaa41838c (2022-11-08)

The names are consistent with that in the commercial PDK release

MOS transistors are: nmos_3p3, pmos_6p0, …
Metal Names are: Metal1, Metal2, …

In newer OpenPDK releases:

MOS transistors are: nfet_03v3, pfet_06v0
Metal Names are: M1, M2, …

Issue

This discrepancy causes some issue when tools from different worlds are used together:

• LVS is problematic because the name differences. The transistor names need to be modified manually.
• The OpenLane Docker image released with GFMPW-0 is nolonger working. I used the older version of the PDK to do digital synthesis.

Link

google doc

Is it some issue in the folder create part, as it will not create the simulation folder by itself

The README file says the following;

Requirements: Python 3.6/3.7 (packages getopt, math, numpy, os, re, shutil, subprocess, sys, smtplib, datetime, logging, matplotlib). Python 3 is not supported.

I assume that last sentence should be Python 2 is not supported?

I was using calibre/2019.3_25 when "make gen_12lp_full" but it didn't generate ldo.pex.netlist which cause the post pex simulation not working. I changed to calibre/2020.3_38.22_aoj and solved the issue.

Hi, please take a look at the src folder, I think it is just a mismatch between the python and the folder files

When I run adc_gen tsmc65lp macro mode, and this is some error, and I have checked the folder, I have also attached the result in the picture

There are some small bugs on the check_env, I am getting errors for checking python version and other cadence tools. And there should be some check for the python libraries like numpy and so on. So just have that updated.

Based on what we have discussed, we will use the lower case letters for our final outputs.
Memory:
So, would you please change "Power" to "power" and AspectRatio to "aspect_ratio" in the mem-gen output json file?
PLL:
would you please change "Fnom_min" to "fnom_min", "Fnom_max" to "Fnom_max" and "aspect ratio" to "aspect_ratio" in the pll-gen output json file?

You seem to have a whole bunch of generators at https://github.com/idea-fasoc/fasoc/tree/master/generators but very little documentation on what the generators do, how types of ranges they work on, how they are used, etc.

Might be a good idea to add a top level README.md to the generators directory and then a bit more information to each of the generators README files.

A lot of the README files seem to have a "This is the readme" at the top of them. IE https://github.com/idea-fasoc/fasoc/tree/master/generators/adc-gen

The title should be the second line.

There is lot of information about potentially alternative flows in a doc at http://j.mp/openeda-diagram

The OpenRoad Project seems like it would also be a potentially good option.

May I access private/cad?

The page @ https://fasoc.engin.umich.edu/ says;

automatically synthesize “correct-by-construction” Verilog descriptions

There doesn't seem to be any definition about what you mean by “correct-by-construction” nor how your guarantee that result? Since it seems to be a core principle, I would have expected it to be clearly described + defined somewhere?

I have pull a new folder and runs the following command, also, I have access to all repos. And I am get this error, please take a look

As I am currently struggling to update config files and generate new Aux-cell using other PDKs, I was wondering if the source of the "AUXCELL_GEN" tool is available somewhere.

Further, I was wondering how could I extend the platform from tsmc65lp to my own PDKs(like tsmc130). And I would appreciate any help.
Best regards.

I have checked the private folder, and there is no hspice there, is that some linkage problem

For memory_gen, running the macro mode, it is unable to find this folder. Please take a look

There is no script folder from the github private view.