NAMD Wiki: NamdOnKNL

  You are encouraged to improve NamdWiki by adding content, correcting errors, or removing spam.

Build and run options for NAMD performance and scaling on Intel Xeon Phi Knight's Landing.

There appears to be an AVX-512 vectorization bug in the Intel 17.0 compiler that breaks the non-KNL-optimized NAMD kernels. As these kernels are used for alchemical free energy and other non-standard simulations that require special non-bonded kernels we strongly recommend sticking to the 16.0 compilers. There is little if any KNL performance benefit from the 17.0 compilers. For this reason the Linux-KNL-icc and CRAY-XC-KNL-intel arch files report an error if an intel/17.0 module is loaded.

See also NamdOnOmniPath.