HLRS shares its expertise with current and potential users of our systems from academia and industry who need advice on HPC technologies and methods.
Consulting by HLRS can address both basic and more sophisticated questions, both within and outside the context of our mentoring and extended user support programs. Our staff can provide recommendations on the following and other key topics in high-performance computing.
Staff in our Numerical Methods & Libraries Department has expertise in optimizing the performance of codes that use numerical methods for scientific simulation. This expertise is enabled by a deep understanding of both numerical methods and their implementation on modern architectures for high-performance computing.
HLRS can answer questions regarding distributed memory, shared memory, and hybrid parallelization paradigms and programming models, as well as other less frequently used models. We especially have expertise in MPI, OpenMP, and GASPI.
In addition to basic parallel programing, HLRS staff can provide insights on a range of other factors that influence performance of parallel applications, including load balancing, graph partitioning, and grid generation.
With the ongoing growth in the amount of data used in computational research, efficient storage access on HPC systems is becoming ever more important for many applications. Although I/O systems are becoming increasingly complex, HLRS can troubleshoot I/O problems and recommend improvements.
HLRS staff use and can provide recommendations concerning profiling methods for assessing key performance features of computationally intensive codes. This includes tests for assessing POP metrics of parallel efficiency as well as I/O and node-level performance.
HLRS staff can assist in selecting and using compilers with the goal of running user algorithms efficiently on HLRS's computing processors and architectures.
We provide support for a variety of commercial and non-commercial tools for detecting bugs, race conditions, and other problems in complex parallel applications.
HLRS facilitates access for our users to a large selection of commonly used software packages for computational fluid dynamics, computational structural mechanics, molecular dynamics, and global systems science. We can provide advice on which software to choose and assist with troubleshooting.
