ECP SOLLVE: Scaling OpenMP With LLVM for Exascale Performance and Portability

PI: Sunita Chandrasekaran

UDEL Students and Software Engineers: Aaron Jarmusch, Aaron Liu, Felipe Cabarcas

Funding Agency: Department of Energy

Duration: 10/01/2020 – 12/31/2023

Collaborating organizations involved in SOLLVE include:
-Brookhaven National Laboratory (BNL)

-University of Delaware

-Argonne National Laboratory

-Lawrence Livermore National Laboratory

-Oak Ridge National Laboratory

-Rice University

-University of Illinois at Urbana-Champaign

These institutions collaborated to enhance OpenMP’s capabilities, ensuring that scientific applications can efficiently utilize the computational power of forthcoming exascale supercomputers like Frontier, Aurora, and Perlmutter.

Project Summary:

The SOLLVE project (Scaling OpenMP with LLVM for Exascale Performance and Portability) is a subproject of the U.S. Department of Energy’s Exascale Computing Project (ECP). Its primary goal is to advance the OpenMP programming model and its implementation within the LLVM compiler infrastructure to meet the demands of exascale computing. This involves enhancing OpenMP to support complex data structures, deep memory hierarchies, and interoperability with other programming models like MPI, ensuring performance portability across diverse hardware architectures, including GPUs from AMD, Intel, and NVIDIA.

The project was led by Sunita Chandrasekaran, a computational scientist at Brookhaven National Laboratory (BNL) and an associate professor at the University of Delaware. She succeeded Barbara Chapman, who previously led the project while at BNL and Stony Brook University.

More on ECP:
https://www.osti.gov/servlets/purl/1809940