91³Ô¹ÏÍø

Definition

SmartScaling is a methodology that uses advanced machine-learning algorithms to reduce the actual simulation runtime, thus mitigating the above challenges in characterizing multiple PVT corners.

With advancements in technology and the push towards smaller nodes from 7nm down to 3nm, designs are expected to work under different modes, with possibly different clock frequencies along with a range of global variations. Thus, designs have to achieve signoff closure for timing and power across an enormous number of process, voltage, and temperature (PVT) corners.

Characterizing this huge number of PVT corners, library teams across the industry face further challenges like high simulation turnaround time, database disk space limits, license server overloads, and hardware costs.


How Does SmartScaling Work?

SmartScaling provides ML-based adaptive library characterization and scaling of libraries for multiple PVTs. It is a new methodology that is built on top of the PrimeTime? scaling engine, which uses existing characterized libraries as anchor PVT corners to then generate multiple additional PVTs instantly.

SmartScaling provides signoff accuracy for various library views ¨C NLDM, NLPM, CCST, CCSN, and LVF.

Two primary methods for SmartScaling are shown below.

Method 1

  • SmartScaling-based PVTs generated via an existing characterized database, used as anchor PVTs.
  • Here, based on the anchor corners provided by the customer, the SmartScaling engine will generate additional libraries at multiple PVT corners.
SmartScaling Method 1 | Synopsys

Method 2

  • SmartScaling-based characterization for overall multiple PVT corners, where the engine will decide which PVT corners to fully characterize and do additional limited sparse corner characterization. This approach essentially creates a database from which to generate additional PVT corners instantly.
  • 20 PVTs = 5 Anchor Char + Sparse Char + 15 SmartScaling PVT corners
SmartScaling Method 2 | Synopsys

The Benefits of SmartScaling

SmartScaling helps to reduce the overall runtime for generating multiple PVT corners for library characterization. Users can:

  • Reduce library characterization runtime by 3x to 10x
  • Reduce library size by 3x to 10x
  • Ensure library accuracy
  • Enjoy zero cost / instant additional PVT corners library generation

Library Characterization and Synopsys

Synopsys offers a comprehensive library characterization solution.

PrimeLib is a unified library characterization and validation solution. The PrimeLib solution includes a comprehensive array of library characterization and quality assurance capabilities that are tuned to produce PrimeTime signoff-quality libraries with maximum throughput on available compute resources. Its innovative technologies utilize embedded gold reference SPICE engines to provide a characterization speed up of advanced Liberty? models used by PrimeTime static timing analysis (STA) to accurately account for effects seen in ultra-low-voltage FinFET processes that impact timing. This includes PrimeTime parametric on-chip variation (POCV), advanced waveform propagation (AWP), and electromigration (EM) analysis. PrimeLib is cloud-ready, and with its optimized scaling technology delivers an accelerated throughput on cloud or an on-premise cluster.

Continue Reading