All users on HPC systems rely on HPC applications for modelling and simulation. Getting the best out of the applications you are using is critical to advancing your research and maximising the value of the HPC resources you have access to.
This page brings together information on:
- Compiling performance versions of popular HPC applications
- Benchmarks and performance data for HPC applications
Compiling HPC Applications
Information on compiling HPC applications is usually available as part of the standard application documentation. However, many HPC applications are complex and the specific instructions for different HPC platforms can differ from the standard instructions.
A number of resources are available with compile instructions for specific UK HPC facilities.
Compiling HPC Applications on ARCHER The ARCHER CSE service at EPCC provides instructions for compiling many applications on the ARCHER UK National Supercomputing service, a Cray XC30 system running CLE 5 with the Cray, Intel and GCC compilers; and Cray MPI.
Compiling HPC Applications on UK facilities The HPC-UK community build repository provides compilation instructions for compiling many applications on UK HPC facilities hosted on GitHub. Contributions to this repository from the community are welcome.
Compiling HPC Applications on Thomas UCL provides compilation scripts for compiling many applications on the Thomas Tier-2 HPC service, a Lenovo HPC cluster running Linux with the Intel and GCC compilers; and Intel MPI and OpenMPI.
Benchmarks and Performance Data
The following links provide more information on benchmarks and performance data on UK HPC systems.
ARCHER Benchmarks The ARCHER CSE service at EPCC have compiled a small number of large benchmarks to test scaling and performance for real-world scientific problems. Both the benchmarks and latest performance results are available along with a white paper describing how the benchamrks were selected.
ARCHER KNL Performance Comparison The ARCHER CSE service at EPCC have compared the performance of Cray compute nodes based on the KNL Intel Xeon Phi manycore processor to standard ARCHER compute nodes based on the Ivy Bridge Intel Xeon CPU. There are results and reports for a range of different HPC applications.