Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a research presentation from USENIX ATC '25 that introduces ATLAS, an innovative attribute-guided approach for discovering compiler bugs through strategic compilation space exploration. Learn how researchers from Nanjing University developed a method that leverages programming language attributes like __attribute__((always_inline)) to insert targeted directives into test programs, enabling more thorough compiler testing than traditional uniform compilation option approaches. Discover how ATLAS provides flexible control over compilation processes by specifying attributes for individual program elements such as functions and variables, allowing investigation of previously unexplored compiler areas. Examine the experimental results demonstrating ATLAS's superiority over baseline testing techniques in bug detection and code coverage across GCC and LLVM compilers. Understand the practical impact of this research through the discovery of 73 unique compiler bugs, with 58 already confirmed or fixed, highlighting the method's effectiveness in uncovering critical issues in foundational system software infrastructure.
