Research Areas
The center organizes its work into seven areas. Each section below describes an area, lists the projects working in it, and links to its publications.
Performance Modeling
Analytical and empirical models that predict how applications and system software behave at scale, and the measurement infrastructure that grounds them.
SparseAtlas is developing a curated and computationally characterized collection of sparse matrices. Its goal is to help researchers understand how sparse-matrix structure influences computational behavior across hardware architectures, algorithms, and scientific workloads.
Solver Design
Numerical methods and parallel solvers built for the communication and memory characteristics of emerging architectures.
System Software
Runtimes, communication libraries, and operating system services that applications depend on for portable performance.
Design, implementation, and evaluation of a stream-triggered MPI interface, initially prototyped and tested on the LLNL Tuolumne and ORNL Frontier HPE Slingshot Systems.
Research to refine and fully specify the syntax and semantics of our stream triggering MPI operations and have them added to the MPI standard as a new “Queued Communications.” The link points to a recent draft of the standard that includes this chapter.
Integration
Bringing the center’s components together into working end-to-end systems, and validating them against production workloads.
Research to refine and fully specify the syntax and semantics of our stream triggering MPI operations and have them added to the MPI standard as a new “Queued Communications.” The link points to a recent draft of the standard that includes this chapter.
Project Infrastructure
The testbeds, build and benchmarking pipelines, and software engineering practice that keep the center’s work reproducible.
A collection of both introductory and advanced training materials about the various applications and software stacks utilized by COMPASS researchers. Also includes a growing collection of hackathon topics.
Education
Curriculum, training, and mentoring that prepare students for research and practice in high performance computing.
A collection of both introductory and advanced training materials about the various applications and software stacks utilized by COMPASS researchers. Also includes a growing collection of hackathon topics.