Collaborative Research: Framework: Software: NSCI: Enzo for the Exascale Era (Enzo-E)
Authors/Creators
- 1. University of California, San Diego
- 2. Columbia University
- 3. Michigan State University
- 4. Georgia Tech
Description
This project is concerned with getting our adaptive mesh refinement (AMR) astrophysics and cosmology application to run efficiently on upcoming Exascale platforms. Our approach has been to develop Enzo-E, the "Extreme" scalable version of our mature ENZO structured-AMR MPI code. Two key features of Enzo-E are 1) it is built on a re-designed scalable "array-of-octree" AMR framework, Cello; and 2) it is parallelized using Charm++ rather than MPI. Charm++'s asynchronous and data-driven task-based parallel programming approach explicitly targets Exascale application development, and is well-suited for the irregular and dynamic data structures inherent in AMR applications. Weak scaling of Enzo-E is excellent, with demonstrated scaling on some of the largest NSF systems available, including Blue Waters at NCSA and Frontera at TACC. Recent accomplishments include additional
physics capabilities, scalable I/O, Version 1.0 release, and the first science publication citing results obtained from Enzo-E simulations.
Notes
Files
2207-cssi.pdf
Files
(3.7 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:aa1a55053fc9047e88c0a0ddddc24e9f
|
3.7 MB | Preview Download |