Pegasus Computational Pattern Analytics

Cadence^® Pegasus^™ Computational Pattern Analytics (CPA) is an advanced high-performance, pattern-based technology suite comprising of mature production-proven flows and engines as well as new innovative approaches. Pegasus CPA includes feature rich capabilities in the pattern-search/classification/profiling domain while leveraging unique pattern engines such as Squish^™ technology, machine-learning (ML), data analysis methods, Python-based scripting, and novel feature-extraction techniques. With Pegasus CPA, design and manufacturing teams can assess layout quality, automate design-quality improvements, accelerate design finishing prior to manufacturing, and develop powerful post-silicon layout analysis. Built-in features allow everyone from beginners to data scientists to leverage the power, capacity, and high performance of Pegasus CPA. Pegasus CPA delivers a richer feature set, more efficient processing capabilties,10-100X faster performance, and larger capacity than traditional DRC-based solutions.

Benefits

High performance

• Optimized algorithms and in-memory operations achieve unmatched processing speeds of up to 100X faster
• Full-chip analysis capability for the largest layout databases in most advanced nodes in hours, not days
• High-productivity CPA Studio GUI integrated in Pegasus Design Review enables faster analysis and dispositioning

Scalable

• Scales linearly with area and number of CPUs
• Easily handles from millions to billions of patterns, markers, and data sets
• Cloud production proven

High capacity

• Efficiently processes large full-chip layouts at the most advanced technology nodes
• Full-chip processing, optimization, data mining, model training, and pattern harvesting
• End-to-end flows have been developed to support up to billions of patterns with optimized CPU and memory usage

Production proven

• In production use at major foundries to screen incoming tapeouts
• Enables urgent and critical foundry production fixes
• Foundation of pattern databases used to gather, monitor, and mitigate yield detractors
• In-design optimizations using foundry kits improve the yield of incoming designs

Versatile applications

• Streamlined design optimization
  • Single-pass, single-deck search and optimization
  • Find exact or similar patterns based on pattern banks or ML models
  • Automatically replace or optimize patterns during design implementation
  • Score patterns based on built-in or customer criteria
  • Increases recommended rules usage and DFM score using foundry kits to statistically improve yield
  • Integrated in Innovus™ Implementation System, Virtuoso^® Layout Suite, and Pegasus Design Review Environment
  • Saves hours with automated DRC fixing
• Interactive layout analysis
  • Interactive pattern capture, editing, search, profiling, and optimization integrated in Pegasus Design Review Environment
  • Allows quick deck development, update, and validation
  • Interactive GUI-driven flows with synchronized layout and results viewing
• Pattern analytics, profiling, and classification
  • Automatically and quickly measures layout dimensions to generate pattern profiling reports and histograms
  • Efficient layout feature extraction is fed into ML engine for powerful layout analytics
  • Identifies most frequent patterns and outliers, and groups similar patterns
• Yield detractor monitoring
  • ML-based hotspot prediction, feature extraction, and Python-based analytics
  • Creation and maintenance of pattern database of yield detractors, including pattern scoring, complexity analysis, and similarity
  • Enables systematic pattern-based dispositioning of yield detractors
• Pattern capture, generation, and enumeration
  • Efficiently captures and compares full chip pattern databases with full coverage
  • Pattern comparison identifies new layout patterns in incoming product
  • Generates and varies patterns of interest within their context used for DRC deck validation and OPC calibration, and tests key generation for process hotspot coverage
• Hierarchy injection
  • Creates hierarchical layout from flat layout or flattened layout
  • Recovers hierarchy from layout that has been flattened
  • Reduced database size and accelerates subsequent processing