ICL Research Profile

CLOVER

Overview

The ECP CLOVER project focused on developing three advanced, vendor-agnostic GPU-accelerated numerical libraries. This project addressed the need for efficient computational tools that are compatible across various GPU platforms, catering to the diverse and evolving landscape of high-performance computing.

Each of the three libraries in CLOVER has a distinct focus: SLATE is dedicated to dense linear algebra computations, offering optimized performance for these tasks on GPU systems. Ginkgo specializes in sparse linear algebra, providing efficient solutions for computations involving sparse matrices. Lastly, heFFTe focuses on multidimensional Fast Fourier Transforms (FFTs), a critical component in many scientific calculations, ensuring accelerated performance on GPUs.

By spearheading the development of these three libraries, CLOVER positioned itself as a cornerstone in the realm of GPU-accelerated distributed computing. This initiative represents a significant stride in the ongoing journey to harness the full potential of GPU technologies, underscoring the importance of adaptable, cross-platform solutions in high-performance computing.