Optimizing Tool Integration Is Essential To Design Success

By James Paris and Armen Asatryan

The relationship between a place and route (P&R) application and the collection of system-on-chip (SoC) design implementation, analysis, and verification methodologies and tools has always been very much a two-way street. The P&R system is the base, if you will, of design implementation—it takes the virtual and makes it physical. However, it is useless without well-defined interfaces to all those other electronic design automation (EDA) tools that comprise a comprehensive, first-class design flow that delivers the required functionality combined with optimal performance.

But there has always been one persistent roadblock in that integration—the different database formats used by the various tool suites in the flow. P&R applications primarily read and write library exchange format (LEF) and design exchange format (DEF) files, while most design verification and optimization tools use the graphic design system (GDS) or open artwork system interchange standard (OASIS) formats. To truly support full integration with P&R tools, any design or verification tool must have the capability to read and write DEF files, either through direct read/write capabilities, or by using database conversion processes to “translate” one format into another.

While the core verification tools, such as those used for design rule checking (DRC) and layout vs. schematic (LVS) verification, have been equipped with these functionalities for some time, there is one set of tools that has lagged behind—the design for manufacturing (DFM) tools. Design companies use DFM tools to analyze design layouts and identify changes and optimizations that can improve yield and reliability. While the use of DFM tools can help shorten design iterations and assist teams in developing tailored solutions for their unique design methodology, historically they have not been well-integrated with P&R environments.

In today’s design environment, the ability to transfer data across the entire design flow is crucial to efficient and “shift-left” design methodologies. However, with only limited database conversion options available, the value of DFM tools to the full design flow was not being fully realized. EDA companies constantly strive to improve and expand tool functionality to provide design companies and foundries with greater value and enhanced performance. Mentor, a Siemens Business, recently incorporated a direct write DEF capability in their Calibre YieldEnhancer tool that provides fast, accurate back-annotation of fill, via, and net objects to P&R applications. Let’s take a look at how it works…

P&R back-annotation

DEF format
The DFM back-annotation process frequently uses incremental DEF format, meaning the input file it provides to the P&R application contains just the objects added to the design database, not the original design data. However, there are some use models that require the DEF output to include the source DEF design data merged with the added objects in a single DEF database. Regardless of these structural differences, using the industry standard DEF format ensures the DFM optimization objects can always be read into any commercial P&R tool.

Flexibility
The mix of design methodologies and tools differs from company to company, or even from team to team. For example, some methodologies require tools to both directly read and write DEF data to and from a P&R database, while other flows use a process that reads OASIS/GDS output files, then converts and writes DEF data for input to the P&R database. That variability means foundry and computer-assisted design (CAD) teams need flexibility to build the best solutions for their users based on team and company requirements. This type of flexibility is essential in EDA tools that support a wide variety of users and flows across the industry.

Flexibility also comes into play depending on the foundry rule deck a team uses. Foundry rule decks designed for fill generation are traditionally set up to read and write OASIS/GDS format. Every design company knows rule #1: never change a fully-qualified rule deck. If an existing flow must be augmented to meet new direct read/write requirements, a team would need to implement a secondary rule deck containing a database conversion step to back-annotate the P&R database with the fill data in DEF format. However, when a new process flow and/or rule deck is being developed, the foundry rule deck could be constructed to include the operations required to directly read LEF/DEF format and back-annotate to P&R in DEF format.

Direct Read/Write DEF

The Calibre Direct Read DEF and Direct Write DEF functionalities enable core Calibre applications to read P&R design data using rule deck operations. Direct Read DEF enables LEF/DEF object and property mapping to Calibre layers, while the Direct Write DEF capability enables Calibre applications to write incremental or full design DEF data directly from Calibre operations. Together, they can be used to build tightly integrated flows for P&R engineers.

The Calibre YieldEnhancer tool can identify and implement automated layout enhancements that can improve yield, such as programmable edge modification to adjust line-ends. The tool’s SmartFill and PowerVia functionalities are used to generate correct-by-construction fill and via objects, based on design inputs from the P&R application. Once these changes are implemented and verified, back-annotation is used to update the P&R design database with the DFM optimizations.

While the Calibre YieldEnhancer tool contains the built-in Direct Read DEF capability to map LEF/DEF objects and properties to Calibre layers for input, designers previously had to invoke a command-line utility to back-annotate fill, via, and metal enhancements in DEF format to commercial P&R tools. This utility had a number of processing and operational limitations that restricted its usefulness and added time to tapeout schedules. Design teams are always looking for a tighter integration throughout their entire design flow, so they needed a faster, more efficient means of directly annotating the P&R database with the DFM design elements in DEF format.

The next-generation Calibre DFM Direct Write DEF process can create a DEF format database without the need for an intermediate GDS/OASIS database conversion. Implemented as a rule deck operation, the Direct Write DEF process provides a straightforward way to define the mapping of the fill, net and via objects to DEF layers. Keywords control what data is selected and how it is back-annotated to DEF format. Direct Write DEF provides process uniformity and eliminates the drawbacks and limitations of the utility. Figure 1 illustrates the basic direct write DEF flow.

Fig. 1: Calibre DFM Direct Write DEF flow.

Net and via back-annotation

For net and via object back-annotation, the Direct Write DEF process requires input objects with annotated net properties (layer and net name), eliminating the need for net information stamping. Attaching properties to net and via objects is typically done by operations that are already in the rule deck. The Direct Read DEF process reads the property value for net and via objects on input layers, and understands to which net each object belongs. Because Direct Write DEF is implemented as a rule deck operation, it enables streamlined data flow within the native Calibre environment, without any need for post-processing of data to generate the DEF output.

The Direct Write DEF operation also eliminates the object processing limitations of the command-line utility by providing several automated capabilities:

• Supports back-annotation of all polygon shapes as metal extensions and via caps
• Recognizes multiple orientations of a single via
• Supports via-array detection to reduce DEF file size. Via array detection minimizes the number of placed via instances by grouping and placing them in via arrays instead.
• Provides automated via repair (figure 2)

Fig. 2: Automated repair of incomplete vias.

Fill back-annotation

For fill objects, the semantics of handling compressed fill data are incorporated in the Direct Write DEF solution, enabling back-annotation of fill data directly from the Calibre YieldEnhancer SmartFill functionality, without the need to generate an intermediate GDS/OASIS file.

Foundry rule decks only generate output in GDS/OASIS format. However, when a design team is using a foundry-qualified rule deck to generate fill data, the Direct Write DEF process can be used to back-annotate this fill data to the P&R application by creating a simple secondary Calibre rule deck with the DFM RDB DEF operation and corresponding mapping of the input fill layers to layers in DEF. While a secondary rule deck is needed, the fact that the Direct Write DEF option is written as a rule deck operation makes the process much easier and faster, and allows designers to view the fill data in their native design environment.

Conclusion

Design companies and foundries constantly modify their design implementation flows to deliver the best possible solution for their operational needs, and the best possible designs for the market. While the use of DFM tools can help shorten design iterations and assist teams in developing tailored solutions for their specific design methodology, they are typically not well-integrated with P&R environments. To enable designers to transfer data from these tools to the P&R environment, database conversions are usually required. In the past, conversion options have been limited, reducing their value to the design flow.

For back-annotation of DFM optimizations to a P&R design database, implementing the Calibre direct write DEF solution into the Calibre YieldEnhancer tool enables Mentor, a Siemens Business to support development of both fully DEF-based flows and hybrid flows, as needed. Features like automatic array detection and via fixing minimize via instances for compact incremental DEF, and avoid dropping incomplete via definitions. The flexibility and functionality provided by the direct write DEF process provides teams with the ability to build the optimal flow based on their needs and requirements, without modification to restricted foundry decks.

EDA companies, like design companies and foundries, need flexibility to remain valuable and effective partners in the design flow process. Constant innovation ensures that new and expanded functionality provides the support, accuracy, and performance needed to get designs to market in a timely and profitable manner.

For more information, download our whitepaper “Optimizing the integration of DFM and P&R.”

Armen Asatryan is an R&D technical lead for the Calibre platform at Mentor, a Siemens business. Prior to joining Mentor, he held a variety of R&D management positions in EDA startups, with a focus on database management for yield analysis tools and flow integration with P&R environments. Asatryan holds a M.S. in applied mathematics with a focus in Computer Science, and a Ph.D. in engineering from the Institute for Informatics and Automation Problems of National Academy of Sciences of Armenia.