草榴社区

PrimeLib? Next-Generation Library Characterization: A Discussion with Moninder Bansal

草榴社区 Editorial Staff

Apr 29, 2021 / 4 min read

Designer’s Digest with PrimeLib? Next-Generation Library Characterization and Validation Solution

We sat down with Moninder Bansal, group director of Research and Development, Digital Design Group at 草榴社区, to learn more about 草榴社区’ unified library characterization and validation solution and how it is solving the challenges faced by the industry, especially at advanced nodes.

Moninder Bansal

Hi Moninder! Thank you for taking time with us today to talk about library characterization. Let us know what is so exciting about the next-generation PrimeLib solution?

Moninder Bansal: PrimeLib is a unified library characterization and validation solution providing an embedded and captive simulator. PrimeLib’s innovative technologies utilize embedded golden reference SPICE engines and integrates golden signoff validation for results that correlate-by-construction in PrimeTime? static timing analysis (STA). 

PrimeLib is a unified library characterization and validation solution, providing signoff-quality libraries. PrimeLib’s innovative technologies utilize embedded golden reference SPICE engines and integrate golden signoff validation for correlate-by-construction results in PrimeTime STA.

草榴社区’ PrimeLib solution, with its advanced machine-learning (ML) technologies accurately accounts for variation effects seen in ultra-low voltage FinFET processes that impact timing. PrimeLib is cloud-ready, and with its optimized scaling technology, delivers an accelerated throughput on cloud or an on-premise cluster. With the increasing demands of characterization for an extensive database of process, voltage and temperature (PVT) conditions, PrimeLib’s SmartScaling reduces the overall characterization turn-around time (TAT) across multiple-PVTs. 

Q: Wow, that is exciting! We would like to hear more about this solution. What changed in the design landscape that drove the need for PrimeLib?

Moninder Bansal: Demand for library characterization multiplies 3X from node to node due to increased PVT corners and advanced modeling requirements like variation modeling, reliability, and aging effects. The on-time availability of quality signoff libraries for advanced process nodes is often a critical bottleneck for chip designers, potentially delaying project schedules. With advanced nodes and smaller devices to meet yield and further margining based on silicon data, there arises a need for late design cycle change in process design kit (PDK), which triggers a full library re-characterization causing a delay in schedules.  

Q: Those are some serious requirements for both accuracy and turnaround time. How does PrimeLib address these challenges, especially those related to the increased compute demand for characterization?

Moninder Bansal: PrimeLib addresses these challenges by advancements in characterization using ML and provides 5X faster LVF characterization, which accounts for 60% of the overall runtime. Also, PrimeLib has superior scalability in its DNA. PrimeLib provides almost linear scaling on cloud and cluster and enables customers to run over 100 thousand jobs in parallel. The PrimeLib solution, with its state-of-the-art scalability, accelerates throughput for library characterization, reducing turnaround time from weeks to days. In addition, PrimeLib’s innovative SmartScaling technology, built on top of 草榴社区 PrimeTime scaling engine, enables Zero-cost and accurate timing, power, noise and LVF library generation for multiple PVT corners based on existing anchor PVTs. 草榴社区’ PrimeLib solution was built on the vision to provide the best-in-class next-generation unified library characterization and validation solution for a wide variety of IPs.

The PrimeLib solution incorporates innovative LVF characterization technologies based on advanced ML techniques for best performance and accuracy. In this domain, PrimeLib offers a flexible and dynamic selection of algorithms to automatically determine whether to use the traditional sensitivity-based approach, or advanced ML algorithms, or High-Sigma Characterization, thereby offering the most accurate solution per user needs. 

 Advances in process technology are increasing the impact of Electro-Migration (EM) on the performance and reliability of designs. At the same time, stress and degradation in the performance of transistors with continued usage over a period of time is another concern. On this front, the cell reliability characterization capabilities of PrimeLib, which include EM characterization and Aging characterization, cater to the automotive industry’s needs.  Additionally, to account for design robustness due to potential process shift, PrimeLib offers a tightly integrated solution with PrimeShield?.

Successful IC design requires high-quality libraries and hence integrated signoff library validation is just as important. PrimeLib provides a comprehensive suite of capabilities for Library QA for timing, noise, EM, power and constraints via integration with 草榴社区’ sign-off tools. Users can benefit from the visual HTML reports and heat maps, GUI-based library-to-library comparison capabilities, among other helpful features, for the critical yet time-consuming validation tasks. 

Q: Looking ahead at the increasing shift towards cloud computing, is PrimeLib a cloud-ready and optimized solution??

Moninder Bansal: Library Characterization is a disk-intensive and highly distributed process. With built-in 草榴社区 Common Libraries for distributed storage and distributed computing, PrimeLib is optimized for NFS traffic and disk usage and facilitates massive parallelization during characterization. And therefore, PrimeLib scales linearly to provide the fastest and most efficient use of available resources.

Powered by PrimeLib’s highly flexible and re-entrant use model with built-in recovery mechanisms along with resource elasticity, users can have a hassle-free experience on cloud. 

PrimeLib has had early collaboration and rapid adoption of key technologies at various key customers. We are excited to have endorsements received from , showcasing the cloud-optimized solution.

Q: PrimeLib overall seems very well positioned to address these challenges. What are the future directions of PrimeLib and something that the users can look forward to?

Moninder Bansal: Within PrimeLib, we are continuously working on expanding our offering to incorporate more complex and challenging Ips, including new models and validation capabilities to offer cutting-edge characterization and validation solutions.

And while new offerings and solutions are underway, we also continue to improve our capabilities further to address the constant need for the fastest turn-around time and accuracy.

PrimeLib’s continued objective is to enable IP developers with a robust, reliable and unified solution for all their characterization and validation needs. 

Q: Thank you so much for giving our readers and an overview of 草榴社区 new and exciting PrimeLib solution. Is there anything else you would like to share?

Moninder Bansal: I am excited about how diverse 草榴社区 product teams have members from diverse cultural backgrounds along with gender parity, bringing in the wealth of innovative ideas, which enable us to develop cutting-edge technologies. We are energized and motivated as a team to continue to lead the characterization landscape.

Continue Reading