Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Emily Shemon Project Name: PROTEUS_2015 Division: NE Project title: PROTEUS Neutronics Analysis Associated funding: DOE NEAMS Program Other Systems: Small NE Linux clusters, Mira (BG/Q) - Catalyst so no official allocation Science: Perform verification calculations for the ANL-developed PROTEUS neutronics code for nuclear reactor calculations. Project description: Code: Primarily using the PROTEUS-SN code, a high fidelity, continuous FEM code modeling neutron transport. The code scales extremely well both weak and strong. Finalist for Gordon Bell Prize when first developed. We need an allotment of LCRC resources for intermediate size jobs which don't fit well on the BG/Q (ALCF) queues. Our typical mode of operation is to use as many cores as possible to reduce runtime, but for small problems (geometrically small), the scaling is limited by the number of finite element vertices. Use of reflective boundary conditions increases the time to solution. Therefore small problems with reflected boundary conditions can use only a 1000 or less processors but runtime can be long (>60 minutes) which doesn't fit well for the BG/Q queues. Our exact problem cases vary year to year depending on requests from our funding source (DOE NEAMS). All problems are run to perform verification and validation tests and are typically space-angle-energy convergence studies which are eventually compared with either continuous energy Monte Carlo (gold standard) or legacy homogenized nodal methods (conventional technique). Industry partnership: n/a Project URL: Current FY Hours Used: undetermined amount New FY Requested allocation: 500000 Q1: 125000 Q2: 125000 Q3: 125000 Q4: 125000 Justification: PROTEUS-SN scales extremely well (beyond the core count of LCRC). Finalist for Gordon Bell Prize in supercomputing. Uses pure MPI. Very little I/O (HDF5 file optional). 90% strong scaling in space and angle parallelism provided 2000+ vertices/processor and 2+ angles/processor are used. Scaling beyond that is not effective. Most problems we run are in the 200,000+ vertex range and 32+ angle range. Storage requirements: 1 TB is fine Thank You, The LCRC Accounts System