TAU
A performance evaluation tool that supports GPUs from multiple vendors for HPC and AI applications
Area: Development tools
CASS members: RAPIDS, STEP
Description
TAU is a versatile performance evaluation toolkit supporting both profiling and tracing modes of measurement. It supports performance evaluation of applications running on CPUs and GPUs and supports runtime-preloading of a Dynamic Shared Object (DSO) that allows users to measure the performance without modifying the source code or build system. It can instrument and rewrite application binaries and DSOs using Dyninst. It supports automatic instrumentation of programs written in Python, C++, C, Fortran, UPC, and Chapel. It supports instrumentation of MPI, CUDA, ROCm, OpenMP, OpenACC, OpenCL, and Kokkos runtime systems. It supports tracking the idle time spent in implicit barriers within MPI collective operations.
Target audience
TAU is intended to be used by application developers, developers of HPC runtime systems, and HPC center staff.