NEMO 3-D Development and Scalability Study
NEMO 3-D Development and Scalability Study |
Sunhee Lee & Gerhard Klimeck |
Objective:
- Run large atomistic electronic structure simulations in minutes
- Represent realistically large structures atomistically
- Utilize available peta-scale computers
Approach
- Improve spatial decomposition parallelism
- Include self-consistent calculations
|
Performance / Impact
- Strong scaling of 3D decomposition => have overcome 1D limitation
- Utilized over 10,000 cores for realistic simulation domain size
- Expecting dramatic reduction of simulation time => 10hours to 20 minutes
- Published in SCIDAC proceedings, 2009, Haley, Lee et al.
|
Powerpoint slide as pdf, or as image below.