Need for a recipe to build highly optimized NAMD binaries for AMD servers and clusters.

Discussion created by kkmtech on Mar 29, 2019

NAMD is a very popular open-source molecular dynamics simulation program in the field of computational biology. Due to its high scalability, NAMD has emerged as a very reliable HPC benchmarking application. It will be really great if AMD can provide a recipe with a specific set of instructions to configure and build optimized binaries for the server-grade processors (like the EPYC series). For the current version of NAMD 2.13, specifics could be things like the most compatible GNU-GCC compiler version and MPI flavour and version where it would be possible for the compiler to recognize the micro-architecture and the instruction set architecture. Other optimizations can include those specific to both floating-point operations and AMD processors (Without causing offense, it would be great if we could have something similar to Intel's recipe to build NAMD:


workstation: https://software.intel.com/en-us/articles/recipe-building-namd-on-intel-xeon-and-intel-xeon-phi-processors-on-a-single-node
cluster: https://software.intel.com/en-us/articles/recipe-building-namd-on-intel-xeon-and-intel-xeon-phi-processors-for-multi-node-runs