Hello, A yearly allocation for the LCRC cluster has been requested with the following updated information: Submitter/PI: Stefan Wild Project Name: UNEDF_Optimization Division: MCS Project title: Optimization of nuclear-energy DFT functionals Associated funding: DOE Scidac UNEDF (Universal Nuclear Energy Density Functional) DOE SciDAC TOPS (Towards Optimal Petascale Solvers) DOE Applied Mathematics Research in Derivative-Free Methods for Complex Simulations DOE SciDAC TOPS (Towards Optimal Petascale Solve Other Systems: 45K CPU-hours on the Cray XT4 (Franklin) at NERSC Science: One of the goals of the UNEDF (Universal Nuclear Energy Density Functional) SciDAC proposal is the improvement of current Density Functional Theory (DFT) functionals. These improvements will give theoreticians better tools for predicting the properties and behavior of atomic nuclei over the entire nuclear table. In this project, an optimization over parameter space (using a newly development derivative-free optimization algorithm) leads to new insights into refining DFT functionals used by the software (HFBTHO), which then requires a new parameter optimization. We plan to continue this work with the next generation of DFT functionals. Our work has generated considerable excitement in the UNEDF project and as a result the request for optimization capabilities is growing. This work is also leading to joint publications in science-based journals. Project description: This DFT parameter optimization is expensive: each simulation of a single nucleus at a particular parameter set using HFBTHO can consume up to .3 CPU-hours. We have recently used OpenMP at the nucleus level, in addition to compiler-based optimization, to bring the wall time per simulation down to 1.5 minutes on a single node of Fusion. Our current optimizations over 13 parameters typically require 200-300 evaluations of about 100 nuclei, which can be done on Fusion in 5-7.5 hours using 100 8-core nodes. This gives us up to 6K CPU-hours for each version of the functional we want to optimize. We note that our current computational needs for a single optimization, 100 nodes for 7.5 hours, yields a significant improvement in wall time over our previous needs of 12 nodes for 90 hours. These optimizations are being done using the derivative-free optimization software POUNDerS developed at Argonne. For prototyping purposes, this software currently requires MATLAB on one master node. Our team of 3 users anticipate doing 40 optimization runs during this project cycle. We also expect to analyze functionals based on more nuclei and depending on more parameters, potentially resulting in further increases to the number of nodes we require. Project URL: http://unedf.org/ Current FY Hours Used: undetermined amount New FY Requested allocation: 240000 Justification: Thank You, The LCRC Accounts System