GPUs have become increasingly prevalent as powerful accelerators for scientific computing applications, offering significant speedup potential over traditional CPU-based implementations. However, achieving optimal performance on GPUs can be challenging, particularly for applications that exhibit irregular memory access patterns or require a high degree of parallelism. In this talk, we present a case study of a computational fluid dynamics application that employs the discontinuous Galerkin method, which poses new challenges for efficient GPU implementation. We investigate novel performance strategies that address these challenges, including the use of shared memory, optimized memory access patterns, and algorithmic optimizations to improve parallelism. Through performance evaluations on GPU architectures, we evaluate the effectiveness of these strategies in improving application performance. The study underscores the need for innovative approaches to fully exploit the potential of GPUs in scientific computing applications.

To add to calendar:

Click on:  https://wordpress.cels.anl.gov/cels-seminars/

Enter your credentials

Search for your seminar 

Click “Add to calendar”