Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Cell libraries (multiple)

Seems that the GDS-format layout tools don't provide the same ability to reference multiple cell libraries in layout. For example there might be a foundry's ESD / I/O library, a "core devices" library, maybe some digital, analog, RF macrocell libs and then the user's collection of art (common cells and mask-set-specific chip layout, which might have several for whatever design management reasons.

I've been evaluating another ($$$) tool that shows this as a significant (to me) limitation.

I'm not seeing anything that looks like it touches on this, in the discussions Search. Maybe I have to read the manual. But I'd be interested to hear from anyone doing full chip design with a high mix of libraries and providers, how what's possible / practical in KLayout compares to what you can do (rack 'em up and pick, ala carte) in commercial tools like Cadence Virtuoso and Silvaco Expert (where they "own" the native format, comprehend the external filesystem and are not constrained by the fact that GDS-II is kinda dumb about external "stuff").

Comments

  • You're right - GDS-based tools don't offer this functionality.

    That's basically because GDS (and OASIS) lacks significant features for providing the infrastructure required for a real design system. I have tried to add a library reference feature in KLayout, but at the cost of compatibility and in a tool specific way.

    There are some old (Magic, Static) and emerging (KiCad) open source alternatives. Maybe you want to look into these.

    Basically I see a demand for a free and open source design system. And I guess it's just a matter of sufficient sponsorship to make this reality.

    Matthias

  • Hi, Matthias,

    I agree about the value of a free design system (curious about what sponsorship might entail). And I am continuing to look at other tools such as MAGIC (whose maintainer, in fact, recommended I keep an eye on your work as an alternative). The libraries limitation is a "bump in the road" for moving designs from commercial tools to open source. Not one that can't be dealt with, but added effort.

    Not familiar with "Static" unless this is StaticFreeSoftware's Electric system.

  • Hi,

    Yes, by "Static" I meant "Static Electric". And it's nice to hear KLayout is being recommended :-)

    And yes, I also feel that library management is not ideal, although I think that with some configuration effort (i.e. setting up a technology) you can basically establish libraries even with GDS and OASIS in KLayout.

    The real missing pieces as compared to a design system IMHO are schematic support (schematic entry/net (back)annotation) and better editing support.

    Kind regards,

    Matthias

  • Hi, Matthias,

    I had continued this message at greater length through the klayout.de messaging system (had more questions / comments that are not for the public necessarily).

    There is a surplus of usable schematic tools open source but few good layout and even fewer verification options. From my perspective if klayout has a good enough DRC (seems to) and a SPICE netlist export (appears to be in work), then other tools can complete the "design system" - maybe just a little effort in formatting things to agree with one or more of the top schematics packages (some are suited to IC design, some less so - have seen some which do not let users add properties and control netlisting, others are more flexible, documentation and examples are always a question).

    For example XCircuit and Netgen (there are multiple projects of that name but one is a good netlist comparison tool) are part of the open source collection maintained by efabless / opencircuitdesign (which includes Magic as well). No need to reinvent those front-end and connectivity / parameter verification tools, only a need to play well with them. They are being used for relatively large IC designs on a commercial (open) basis today. Ability to read back a Netgen error report and highlight what it calls out, would be a big deal for usability but much, much less effort than trying to write your own LVS.

  • Well, if I had thought that way, I had never started KLayout. There are enough GDS editors and viewers out there.

    Do worry about effort I'm spending. LVS is already done.

    Matthias

  • There may be "enough editors" but there are not enough good ones. I like the behavior of klayout better than others I've tried (and paid serious money for).

    Can't wait to see the LVS in action.

  • For a first version you can look here: https://github.com/KLayout/klayout/wiki/2019-04-14

    Downloads for the builds are available here: https://www.klayout.org/downloads/dvb/

    Currently lacking: a nice wrapper language like the DRC script, a more elaborate Spice reader for the reference netlist and some more advanced netlist normalization features such as blackboxing.

    I'm trying to keep a blog about the progress here: https://github.com/KLayout/klayout/wiki/Deep-Verification-Base

    Matthias

Sign In or Register to comment.