Advanced Platform Technology Center
High Performance Computing Facility
Located at LSCVAMC
The High-Performance Computing Facility at the LSCVAMC is a local computational resource for musculoskeletal modeling, biomechanical simulation and controller design. Directed by Dr. Musa Audu, this is a state-of-the-art computational facility that serves as a central resource for the high performance computing requirements of several biomechanical researchers associated with the APT Center. This facility is equipped with state of the art super-computing facilities such as a 16-processor supercomputer - Fusion A8 server with 8 dual core processors and two new super-computers (HPX QS8-8500) each running with 32 AMD Opteron processors. Put together, these give us a cluster with a total of 72 processors that can be run in parallel, suitable for all kinds of rapid dynamic optimization studies. Local software resources resident on these computers include SIMM (Software for Interactive Musculoskeletal Modeling) and SD/FAST, Matlab 2013 and Microsoft Visual Studio 2010. Others are the open-source software Pseudospectral Optimal Control Solver in C++ (PSOPT) (developed by Victor Becerra at University of Reading, UK) and Asynchronous Parallel Pattern Search Package (APPSPACK) developed at Sandia National Laboratories. These are all powerful, state-of-the-art software for advanced 3D musculoskeletal modeling.
The computational resources at the Ohio Supercomputer Center (OSC) and the Case High Performance Computing Cluster (Case HPC) are also available for the modeling, simulation and design work. Both facilities provide parallel computing facilities that allow the solution of a wide variety of optimization problems that would otherwise be difficult with stand-alone computers. These clusters, and our own local cluster, are all based on the open-source Linux operating system and offer a comprehensive array of parallel software such as Intel C++, FORTRAN, and Matlab. Also available are specialized analysis packages, along with powerful libraries such as the Intel Math Kernel Library (IMK), which include the most versatile linear algebra libraries such as LAPACK and BLAS.
These computational resources are critical to the solution of a wide variety of problems in musculoskeletal modeling involving a large number of rigid segments and muscle elements such as those in our musculoskeletal model.