Lattice QCD on Xeon Phi
We will use Tango to develop code and run benchmarks of the MILC code suite on Intel KNL processors. We are part of the Indiana University Intel Parallel Computing Center. We will compare results on Tango to systems with other networks and versions of the processor.
Use of FutureSystems
Code development and benchmarking
Scale of Use
Run scaling tests of code on tango. May need to benchmark multiple versions of code. Scaling test could take 10s of node hours. If the system is not heavily utilized, could try an actual physics project.