Compiling-for-performance checklist on Ookami

For getting good performance on Fujitsu A64FX there are a few things to consider during compilation. Please have a look at the getting-started guide and also see the checklist below for guidance:

• Compiler(s)
  • Version(s)
• Optimization flags
• Vectorization flags
• Introspection flags (ask compiler to tell you what it is optimizing/vectorizing)
• Is the code parallelized?
  • How?
    • MPI + version + underlying compilers
    • OpenMP
    • Threads
    • Other
• Layout of code in NUMA regions, threads, and across compute nodes
  • MPI ranks 
  • Threads per rank
  • Placement of ranks

We put together a matrix of useful options for our various compilers here 

If you have any questions please use the slack channel, the office hours, or submit a ticket