Scalable Solvers Lab
Research Goals
Parallel computing is everywhere! Your personal computer likely has multiple cores, and possibly multiple CPUs. For programs to run efficiently, both with respect to time and energy, parallel computing is essential. Large scale simulations, numerical solvers, and large language models are making discoveries and scientific impacts because of the vast amount of parallel computing power available today. Emerging supercomputers have reached exascale performance, meaning they can compute one million-million operations per second through the use of GPU accelerators. However, applications are often unable to take full advantage of available compute power due to the high costs of inter-process communication. The Scalable Solvers Lab at UNM is creating portable optimizations for efficient communication, improving collective operations, creating portable MPI extensions, benchmarking emerging systems, and creating tangible improvements to widely used parallel applications and solvers in an effort to have applications catch up to hardware achievements and reach exascale performance.
