Thursday, February 24
MS44
Testing Numerical Code for Heterogeneity: Lessons Learned and Emerging Projects
3:20 PM - 5:00 PM
As applications in the domains of scientific computing and machine learning continue to transition to more complex heterogeneous environments, testing such numerical codes continues to be a challenge. While testing numerical code on CPU systems has been a standard practice for years, traditional methods and tools do not automatically apply to heterogeneous systems—compiler optimizations, high degree of asynchrony in GPUs, emerging performance portability models, and reduced floating-point precision are just some of the challenges that testing software faces. In this minisymposium, we bring together experts from different fields to present some emerging projects on testing numerical code as well as lessons learned in existing scientific software.
Organizer:
Ignacio Laguna
Lawrence Livermore National Laboratory, U.S.
Ganesh Gopalakrishnan
University of Utah, U.S.
Ang Li
Pacific Northwest National Laboratory, U.S.
Pavel Panchekha
University of Utah, U.S.
Cindy Rubio González
University of California, Davis, U.S.
Zachary Tatlock
University of Washington, Seattle, U.S.
-
3:20-3:40 ComPort: Rigorous Testing Methods to Safeguard Numerics During Software Porting
abstract
- Ganesh Gopalakrishnan, University of Utah, U.S.; Ignacio Laguna, Lawrence Livermore National Laboratory, U.S.; Ang Li, Pacific Northwest National Laboratory, U.S.; Pavel Panchekha, University of Utah, U.S.; Cindy Rubio González, University of California, Davis, U.S.; Zachary Tatlock, University of Washington, Seattle, U.S.
-
3:45-4:05 Toward Automated Test Generation for Performance Portable Programs using Clang/llvm and Formal Methods
abstract
- Shyamali Mukherjee and
Noah Evans, Sandia National Laboratories, U.S.; Alessandro Orso and
Vivek Sarkar, Georgia Institute of Technology, U.S.; Keita Teranishi, Sandia National Laboratories, U.S.
-
4:10-4:30 Testing Strategies in a Large Scale Scientific Code – Experiences and Aspirations
abstract
- Balint Joo, Oak Ridge National Laboratory, U.S.
-
4:35-4:55 Testing, Verifying, and Validating Openmp Implementations
abstract
- Swaroop Pophale, Oak Ridge National Laboratory, U.S.; Nolan Baker, University of Delaware, U.S.; David E. Bernholdt, Oak Ridge National Laboratory, U.S.; Sunita Chandrasekaran,
Josh Davis,
Thomas Huber, and
Sergio Pino Gallardo, University of Delaware, U.S.