Between The Lines

Oct 26 2007 4:01PM | Permalink | Email this | Comments (6) |
Blog This! using:  Blogger.com | LiveJournal |
Digg This | Slashdot This | add to Del.icio.us

Hi folks, just months after its founding, the Interoperable PCell Libraries (IPL) coalition is showing great progress, as this week at the Synopsys Interoperability developer’s forum the group demonstrated a multi-vendor analog/AMS tool flow running IPL’s newly released library on the Si2’s OpenAccess common database.

In particular, IPL members demonstrated Silicon Navigator’s RDE Framework, Ciranova’s Pyros layout viewer, Synopsys Hercules PVS (DRC/LVS tool), Silicon Canvas’ Laker analog layout tool, and Mentor Graphic’s Calibre DRC/LVS working together in a bidirectional flow on OpenAccess.

Dave Millman, vice p of marketing over at Ciranova, said traditionally when users wanted to transfer a design from one vendor’s tool to another vendor’s tool, the tool receiving the design would traditionally have to translate the data to its own format to read it. Doing that can be time consuming and in some cases important data can literally get lost in translation. To make matters worse some of the translators only translate in one direction, which means if a change is needed, you have to do a lot of manual work to amend the original file with fixes. But IPL members have now demonstrated several vendor tools seemingly seamlessly sharing the same Pycell library data via the Open Access common database. And it’s a bidirectional flow—meaning any changes to a design are instantly stored in OA and are accessible by other tools connected to OA.

To refresh your memory, the IPL was started by a bunch of Cadence competitors in response to Cadence not making its proprietary Skill language available to competitors. Skill is used among other things to describe analog cells called Pcells and foundries typically develop their Process Development kits, specifically the libraries of their latest and greatest processes in Skill-based Pcells. Cadence has used its proprietary language to maintain its two decade dominance in analog IC layout, as its tools are the only ones that can read Pcells without major problems.

Because they don’t have access to Skill for Pcell development, IPL members are backing a Ciranova developed Pcell description language based on Python and Ciranova has made its PyCell studio tool for developing Pycells available to anyone free of charge. To boot, Ciranova and the other members maintain that Python is a much more modern and elegant language than Skill and that Pycells are superior. 

However, this latest demonstration by IPL is just one of many steps IPL needs to make to achieve its goal of analog and AMS tool interoperability on OA.

Thus far the IPL has created a Pycell library and has shown the library can be accessed by OA compliant tools. The next step toward making a completely interoperable flow is to create a common set of properties, parameters and constraints for Pycell library developers and the way tools access those libraries. To do this, IPL members have now established three working groups. The first is chaired by Ciranova’s Ed Petrus and works on Pcell development. The second will work on Properties and Parameters and is chaired by Synopsys’ Neel Gopalan, while the third working group, also headed by Petrus, is developing constraints.

AWR, Ciranova, Magma, Silicon Canvas, Silicon Navigator and Synopsys all have members in these working groups.

If and when, it gets that work done, it then must convince foundries to create their newest PDKs in Pycells. Incidentally, IPL members claim Cadence tools can also work with the other IPL member tools via the OA common database. Ciranova offers a tool that essentially allows OA tools to read PCells. So perhaps the biggest feather in IPLs cap is to get Cadence to participate in the discussion, let alone participate in the development of IPL.

Incidentally, I also heard a rumor that Cadence may be lightening up a bit or is perhaps a bit worried and has started to make Skill licenses available. I haven’t been able to confirm this with Cadence, yet.

Related entries in: EDA |