OpenYield: An Open-Source SRAM Yield Analysis and Optimization Benchmark Suite

2025-08-06

AI Summary

SRAM yield analysis suffers from a disconnect between academically simplified models and industrial reality, compounded by the absence of open, reproducible benchmarks—hindering practical adoption of research. To address this, we introduce OpenYield, the first open-source, industry-aligned benchmark platform for SRAM yield modeling. It features a high-fidelity circuit generator incorporating secondary parasitic effects, inter-cell leakage coupling, and peripheral circuit process variations. OpenYield provides a unified interface supporting SPICE-level simulation, baseline yield evaluation, and robustness-aware optimization algorithms. It further delivers standardized evaluation protocols and automated design-analysis capabilities. By enabling rigorous, reproducible, and comparable yield studies, OpenYield bridges the academia–industry gap and accelerates collaborative innovation and engineering deployment of high-assurance SRAM reliability techniques.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Static Random-Access Memory (SRAM) yield analysis is essential for semiconductor innovation, yet research progress faces a critical challenge: the significant disconnect between simplified academic models and complex industrial realities. The absence of open, realistic benchmarks has created a reproducibility crisis, where promising academic techniques often fail to translate to industrial practice. We present extit{OpenYield}, a comprehensive open-source ecosystem designed to address this critical gap through three core contributions: (1) A realistic SRAM circuit generator that uniquely incorporates critical second-order-effect parasitics, inter-cell leakage coupling, and peripheral circuit variations, which are typically omitted in academic studies but decisive in industrial designs. (2) A standardized evaluation platform with a simple interface and implemented baseline yield analysis algorithms, enabling fair comparisons and reproducible research. (3) A standardized SRAM optimization platform, demonstrating OpenYield's utility in enhancing SRAM design robustness and efficiency, providing a comprehensive benchmark for optimization algorithms. OpenYield creates a foundation for meaningful academia-industry collaboration, accelerating innovation in memory design. The framework is publicly available on href{https://github.com/ShenShan123/OpenYield}{OpenYield:URL}

Problem

• Bridging gap between academic SRAM models and industrial realities
• Addressing reproducibility crisis in SRAM yield analysis research
• Providing open benchmarks for SRAM optimization algorithm evaluation

Innovation

• Realistic SRAM circuit generator with second-order effects
• Standardized evaluation platform for reproducible research
• Comprehensive benchmark for SRAM optimization algorithms