Project Description

KUDA is a framework for GPU accelerated dynamic program analyses.

KUDA currently includes data-race detection (the Eraser and Goldilocks algorithm) and reduction-based atomicity-violation detection algorithms. We are planning to extend it with deadlock and livelock detection algorithms.

KUDA is developed at the Koç University Research Center for Multi-Core Software Engineering at Koc University (where KU comes from).

For more information, see our WoDet3 paper:
KUDA: GPU accelerated split race-checker framework
or our technical report
Rethinking Runtime Verification on Hundreds of Cores: Challenges and Opportunities.

People:

Tayfun Elmas
The Parallel Computing Lab. (ParLab), UC Berkeley
http://sites.google.com/site/tayfunelmas/

Can Bekar
The CADT Lab., Koc University, Istanbul, Turkey
http://canbekar.com/

Semih Okur
The CADT Lab., Koc University, Istanbul, Turkey
http://www.semihokur.com/

Serdar Tasiran
The CADT Lab., Koc University, Istanbul, Turkey
http://home.ku.edu.tr/~stasiran/