ICL Research Profile

Scalable Performance and Accuracy analysis for Distributed and Extreme-scale systems

Overview

ICL is breaking new ground with SPADE, an NSF-funded "Frameworks" project titled "Scalable Performance and Accuracy analysis for Distributed and Extreme-scale systems (SPADE)." ICL leads this collaborative project with Heike Jagode as the PI and Anthony Danalis as co-PI. The project also involves partnerships with the University of Maine (co-PI: Vince Weaver) and the University of Texas, El Paso (co-PIs: Shirley Moore and Christoph Lauter). Spanning four years, the project started on September 15th, 2023, with a total budget of $3.5M (ICL's share is $2.1M).

The SPADE project is dedicated to enhancing monitoring, optimization, evaluation, and decision-making functions for extreme-scale systems, catering to the needs of the High-Performance Computing (HPC) and scientific applications communities. As HPC resources evolve towards extreme scale, there's a growing necessity for integrated frameworks to tackle performance and reliability issues. Through extending support for heterogeneity and scalability across various computing platforms, and employing the established PAPI performance monitoring library, SPADE aims to provide the necessary software and APIs to effectively address the demands of scientific and machine learning applications while exploring new accuracy versus performance trade-offs with low-precision floating-point types.