This research introduces a Retrieval-Augmented Generation (RAG) approach to automate compiler testing across different hardware architectures, significantly improving efficiency and coverage.

• Enables automatic generation of test cases for OneAPI's cross-architecture compilers without requiring deep compiler expertise
• Leverages AI to target specific compiler paths and optimizations that traditional fuzzing cannot reach
• Demonstrates superior code coverage compared to conventional testing methods
• Reduces the time and expertise needed to verify compiler correctness, reliability, and security

For engineering teams, this innovation means faster development cycles and more reliable cross-platform software with reduced security vulnerabilities - critical for organizations deploying across diverse hardware environments.

RAG-Based Fuzzing of Cross-Architecture Compilers