Profile Util library: A quick and easy way to get MPI, OpenMP and GPU runtime information
Pawsey Supercomputing Research Centre, Kensington, WA, Australia
EuroMPI Australia, 2024
@article{elahi2024profile,
title={Profile Util library: A quick and easy way to get MPI, OpenMP and GPU runtime information},
author={Elahi, Pascal Jahan},
year={2024}
}
We present profile_util, a quick and simple way of profiling codes. This is a MPI, OpenMP, and GPU enabled C++17 library. The GPU interface is compatible with both HIP and CUDA and is compatible with more than a single GPU per MPI process. It provides a means of logging MPI, OpenMP and GPU related information, such as number of threads, core affinity, GPU properties. It can also be used to log memory usage or available memory. The library can also report the time taken on both host and devices, statistics of CPU usage, statistics of GPU usage, memory usage and power. The simple API means minimal changes are required to any C++ library.
November 10, 2024 by hgpu
