Abstract:Convolutional neural networks (CNNs) are widely adopted in safety-critical systems, including space applications and autonomous vehicles. Field-programmable gate arrays (FPGAs) based on SRAM are preferred for accelerating CNN computations due to their unique characteristics. However, the configuration memory of FPGAs is susceptible to single event effects (SEEs), which can corrupt computations and lead to misclassification of CNN outputs. In this study, we investigated the impact of SEEs on SRAM-based FPGAs with Two-Photon Absorption (TPA) laser fault injections through a comparative analysis of two popular CNN acceleration architectures: streaming architecture (SA) and single compute engine (SCE). Experimental results show that SA-based CNNs require more hardware resources but exhibit superior resilience against single event upsets (SEUs). Without any Radiation Hardened by Design (RHBD) protection, SCE has an error rate approximately twice as high as SA. To mitigate errors, the Xilinx IP core - Soft Error Mitigation (SEM) is used for error detection and correction, leading to error rate reductions of up to 50 % in both architectures. Importantly, we propose the AutoDPR-SEM (Autonomous Dynamic Partial Reconfiguration for Soft Error Mitigation) approach, which automatically reconfigures the SEM IP core when it remains idle due to uncorrectable errors. AutoDPR-SEM significantly improves CNN error rates, reducing errors by approximately 17.8 times in SCE and 14.8 times in SA. We also applied software level simulation to validate the TPA experiment, showing similar trends of the testing results across all models. In conclusion, the study confirms the feasibility of AutoDPR-SEM in both architectures, showcasing its potential to improve CNN error rates in safety-critical systems.

Design Rule Checking with a CNN Based Feature Extractor

Efficient Design Rule Checking Script Generation Via Key Information Extraction

DRC-Coder: Automated DRC Checker Code Generation Using LLM Autonomous Agent

Effective Fabric Defect Detection Model for High-Resolution Images

Accelerating Chip Design with Machine Learning: from Pre-Silicon to Post-Silicon

Hierarchical Design Rule Checking Approach for IC Layout

An Efficient Approach for DRC Hotspot Prediction with Convolutional Neural Network.

A Smart Surface Inspection System Using Faster R-CNN in Cloud-Edge Computing Environment

In Design DFM Rule Scoring and Fixing Method using ICV

Comparative study: AutoDPR-SEM for enhancing CNN reliability in SRAM-based FPGAs through autonomous reconfiguration

Transforming Global Routing Report into DRC Violation Map with Convolutional Neural Network

DRC Hotspot Prediction at Sub-10nm Process Nodes Using Customized Convolutional Network

Context-Aware DFM Rule Analysis and Scoring Using Machine Learning

ICML: Machine Learning-based Transistor-level Integrated Circuit Layout Error Classification using Color Analysis and Segmentation

Model Checking on Clock Domain Crossing Design of System-on-Chip

DRC-aware Pin Access Planning for Detailed Routing

Automatic Routability Predictor Development Using Neural Architecture Search

An Automated System for Checking Lithography Friendliness of Standard Cells

PCBNet: A Lightweight Convolutional Neural Network for Defect Inspection in Surface Mount Technology

Specification-Driven Automated Conformance Checking for Virtual Prototype and Post-Silicon Designs

Defect Detection of Mechanical Design Products Using Deep Neural Network