Scientific and Public Contributions using High-Performance Computing

We are conducting R&D related to advanced applications of high-performance computing on the TSUBAME supercomputer and aiming at scientific outcomes and public contributions.

• New algorithms for GPU computing

• Large-scale HPC applications using GPU

• International research collaboration on HPC applications

• Dynamic load balance

• AMR (Adaptive Mesh Refinement)

• Hierarchical methods for Exascale

• Scalable deep learning on thousands of GPUs

• International collaboration for open source libraries

Applications on TSUBAME

• Large-scale simulation for two-phase flows

• Large-Eddy Simulation on turbulent flows

• Study on Fluid-structure interaction

• Tsunami simulation

• Particle simulation for granular and fluid

• Hierarchical low-rank approximation

• Kronecker factorization

• Hierarchical communication for deep learning

• Second order optimization for deep learning