Skip to content

UM-ARM-Lab/gpu_profiling

Repository files navigation

ARM Lab GPU Profiling

This repo houses the code the UMich ARM Lab uses to benchmark their GPUs.

Installation

Hardware

These benchmarks require a CUDA-capable GPU, meaning a somewhat recent NVIDIA GPU and driver.

Software Dependencies

The only software dependencies for running these benchmarks is Python 3.7+ and pipenv. Python 3 should be on your path as python3. Your system installation of Python is sufficient as the benchmarking functions handle virtual environment creation for you. Additionally, pipenv should be installed and on your path.

If pipenv is not installed, install it with python3 -m pip install pipenv.

Installation And Initialization Process

Initialize the repository with:

./init_repo.sh

The script clones the necessary repositories and creates a pipenv environment for running the benchmarks. Note that this will ask you which CUDA version you would like for PyTorch to be installed with.

Usage

Running Benchmarks

  1. Ensure that you have initialized the repository by following instructions in the Installation section.
  2. cd to where this repository is located.
  3. Activate the pipenv environment with pipenv shell.
  4. Run the benchmarks with python run_benchmarks.py.

Comparison With Other Rigs

To compare the results of your machine with others, run python compare_results.py.

We include the profiling results of a few of our older computers for reference.

However, the comparison script isn't all-encompassing so if you want fine-grained control of plotting, you may desire to make your own script or notebook.

About

GPU Profiling Tools For The Lab

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors