HPCToolkit

HPCToolkit is an integrated suite of tools for measurement and analysis of program performance on computers ranging from multicore desktop systems to GPU-accelerated supercomputers. By using statistical sampling of timers and hardware performance counters on CPUs, HPCToolkit collects accurate measurements of a program’s CPU work, resource consumption, and inefficiency and attributes them to the full calling context in which they occur. By monitoring GPU operations, gathering instruction-level metrics within GPU kernels, and attributing the costs of GPU work to heterogeneous calling contexts. HPCToolkit provides insight into the performance of GPU-accelerated codes. HPCToolkit works with multilingual, fully optimized applications that are statically or dynamically linked. HPCToolkit is designed for use on large parallel systems. HPCToolkit’s presentation tools enable rapid analysis of a program’s execution costs, inefficiency, and scaling characteristics both within and across nodes of a parallel system. HPCToolkit supports measurement and analysis of serial codes, threaded codes (e.g. pthreads, OpenMP), MPI, and hybrid (MPI+threads) parallel codes, as well as GPU-accelerated codes that offload computation to AMD, Intel, or NVIDIA GPUs.