An open-source performance analysis and profiling tool for parallel programming languages including C, C++, Fortran, Python, and Java, helping developers identify performance bottlenecks in HPC applications.
Tau Analyzer is an open-source, portable, parallel performance analysis and profiling toolkit designed for tracing, instrumentation, measurement, analysis and visualization of high-performance computing (HPC) applications written in C, C++, Fortran, Python and Java programming languages. It helps developers identify performance bottlenecks in their parallel applications running on supercomputers, clusters, and workstations.
Tau can gather performance data through instrumentation of functions, methods, basic blocks, and statements. It supports user-defined events and supports gathering performance data from accelerators and co-processors (GPUs and MIC). The gathered performance data includes profiles, traces, aggregated statistical profiles, and snapshots. It can analyze the performance of threaded, OpenMP, MPI, hybrid, and CUDA/OpenCL codes.
Key features of Tau include very low overhead instrumentation, scalable trace aggregation, integrated sampling, polymorphic event definitions, source code annotation, dynamic instrumentation, flexible I/O support, dynamic runtime layer for portability, interactive graphical user interface for analysis, and cross-platform support.
Here are some alternatives to Tau Analyzer:
Suggest an alternative ❐