HPC DEBUGGING SOFTWARE

Many of the scalable codes we use are based on MPI, Open-MP or both, or even just using multi-threaded single nodes, and will run on many different architectures (CPUs from ARM and Intel, GPUs from nVidia and AMD, etc) with many different storage systems (e.g. Lustre, BeeGFS, Gluster, CEPH) and on many nodes on an HPC system. Some MPI based applications scale up to thousands of cores and dozens of nodes. Ideally we would want to be able scale to this level but the initial ambition is to be able to run the software...