SiCortex

[videos and images]

Visualizing a Kautz Graph.

The SiCortex at Argonne supports research in the Mathematics and Computer Science Division. Such research involves research and development targeting multiple aspects of scalable high-performance computing, especially languages, libraries, and system software. Some of these projects work closely with application development groups in order to motivate and validate their contribution The following is a list of research projects (some still in a preliminary state) and associated applications currently using the SiCortex at Argonne.

MPICH2 -- Portable MPI implementation
ADLB -- Asynchronous Dynamic Load Balancing Library
GFMC -- Green's Function Monte Carlo ab initio nuclear theory code
NEK -- Fluid dynamics
PETSc -- PDE solver library
FLASH -- Astrophysics for Type Ia supernovae
MPIBlast -- Biological Sequence Matching
ParaMEDIC -- Framework for large-scale distributed computing
falkon -- Globus task farming library
Swift -- Grid Workflow software
PVFS2 -- Parallel Virtual File System
PnetCDF -- parallel implementation of setCDF I/O library
Hydra -- next-generation porcess management system for MPICH2
Autodock -- computational biology
mx -- Myrinet communication library development

Our system has: 972 nodes

- 6 cores per node
- 4 GB/memory per node
- 1300 MB/s interconnect bandwidth per node
- 1 us of latency
- the system has a novel network topology, described at:

http://en.wikipedia.org/wiki/Kautz_graph