Yebin Chon will present her General Exam "Augmenting Alias Analysis With An Enhanced KillFlow Algorithm" on Wednesday, October 11th, 2023 at 3pm in Friend Center 202.
The members of her committee are as follows: David August (adviser), Zachary Kincaid, and Amit Levy
Title "A Unified API for Memory Dependence Analysis Tools"
Abstract:
Stronger memory dependence analysis creates additional and more efficient options for automatic parallelization.
CAF, the collaborative analysis framework, developed at Princeton, is to the best of our knowledge the strongest open source memory analysis engine for automatic parallelization of general purpose programs. Using memory profiling, we identify areas of weakness in the CAF memory dependence analysis engine.
Despite CAF’s precision, our investigation reveals that its KillFlow module, while effective in many scenarios, faces limitations when dealing with certain loop patterns found in many applications. We propose a stronger KillFlow algorithm that is able to handle dependences involving inner loops, and show how it disproves many dependences not already disproven by CAF.
Links to Abstract and Reading List:
https://docs.google.com/document/d/1Sd5IqojKpFB24PLRgmyqdrxPRrmodts-0TF_w2KUQwQ/edit?usp=sharing
https://docs.google.com/document/d/1m0vGIihX2A60Z1xRnkTnH7XxSWf7Wn669k9A5tbchG4/edit?usp=sharing