Is it a requirement to provide source files for a ...
# shuttle-precheck
m
Is it a requirement to provide source files for a project on the open MPW? This project from MPW8 does not: https://platform.efabless.com/projects/1881 It looks like they just uploaded a GDS file. Edit: Maybe it wasn't selected for the MPW8?
@Boris Murmann This says it was part of the PICO project? Does that require it to be open source?
b
Yes, we have that requirement, but this design was not finished or taped out.
m
Got it. Thanks for the quick response.
We were looking at all of the ReRAM arrays taped out to see the current status and saw that one.
m
@Matthew Guthaus The requirements for Google MPW state • The project must be fully open. The project must contain a GDSII layout, which must be reproducible from source contained in the project. However, there is no check for source files in the analog precheck and only a top level check in the digital precheck. Hopefully soon we are planning to add LVS to the precheck.
m
Yes, I was mostly wondering if they actually get checked. Thanks for the added info.
t
We are now starting an initiative to specify exactly what needs to be submitted for a successful analog design precheck. The general underlying idea is that any set of tools can be used for the design, but that the design GDS must be submitted along with a valid netlist, and that it must be possible to extract the GDS using Magic and run LVS against the netlist using Netgen and generate a correct LVS result. That will put a few constraints on the way analog projects can be made, mostly eliminating things that I consider to be bad design practices anyway, like splitting devices between subcells, that require the best commercial tools to bend over backwards to sort out and make a valid netlist from. The result, though, will be that for any submitted design GDS, anyone can use open source tools to simulate the design or otherwise be able to reuse it. It would be preferable to require a valid schematic, but that's rather more problematic to enforce.
👍 1
m
That makes sense. While using commercial tools should be ok to make open IP, it really is a focus on open source.
t
And more to the point, it's a focus on re-use, by people who may not have access to the same tools used by the original designer.
t
@Tim Edwards Forgive me if I’m wrong. But as a follow-up on a discussion we had a few weeks back, I realized in the meantime, that the netlist you check for is just the top level (hierarchical) netlist, not the flat one. So you check for a wrapper version with just another syntax than the RTL version. Sorry if I’m wrong, it’s been a while since my submission.
t
@Tobias Strauch: Yes, I believe there is some existing netlist sanity check that is essentially useless (at best). I have no intention of keeping that check, but instead do a proper LVS in a way that is both meaningful and not too much of a burden on users who might have already done LVS but in a different flow using different tools.
t
Sure, but the check I'm refering to was added for 7 or 8, with the intension you kindly explained here: https://open-source-silicon.slack.com/archives/C017HPHCMEY/p1670278280743899?thread_ts=1670113972.314549&cid=C017HPHCMEY I just wanted to rise a flag that a hierarchical netlist version of the wrapper might not be what you were looking for. (Sorry again if I'm wrong.) Cheers.