A Computing Platform for Parallel Sparse Matrix Computations
Technical Report,28 Aug 2012,27 Aug 2013
Purdue University West Lafayette United States
Pagination or Media Count:
This grant enabled the purchase of an Intel multiprocessor consisting of eight multicore nodes interconnected via an infiniband. Each node contains 24 cores. This parallel computing platform has been used by my research group in the early stages of developing large sparse linear system and symmetric eigenvalue problem solvers ARO grant W911NF-07-R-0003-04 that are suitable for parallel architectures containing hundreds of multicore nodes thousands of cores. Once our parallel solvers obtain the correct solutions, and perform properly on this 8-node platform, the codes are ported to much larger parallel computing platforms at the Intel Corporation. Two Ph.D. dissertations have resulted from the above grant awarded by Dr. Myers, and this DURIP grant awarded by Dr. Coyle.