https://github.com/rugleb/cuda

A simple example of a program that uses parallel GPU computing on an NVIDIA graphics card using CUDA technology
https://github.com/rugleb/cuda

cuda gpu nvidia

Last synced: about 1 year ago
JSON representation

A simple example of a program that uses parallel GPU computing on an NVIDIA graphics card using CUDA technology

• Host: GitHub
• URL: https://github.com/rugleb/cuda
• Owner: rugleb
• License: mit
• Created: 2018-11-25T19:38:30.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2019-05-27T19:30:09.000Z (about 7 years ago)
• Last Synced: 2025-02-15T16:49:23.101Z (over 1 year ago)
• Topics: cuda, gpu, nvidia
• Language: Cuda
• Homepage: https://github.com/rugleb/cuda
• Size: 21.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
[![Build Status](https://travis-ci.com/rugleb/cuda.svg?branch=master)](https://travis-ci.com/rugleb/cuda)

[![Language](https://img.shields.io/badge/Lang-CUDA-green.svg)]()

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

A simple program that uses parallel GPU computing on an NVIDIA video card using [CUDA](https://developer.nvidia.com/cuda-zone) technology.

The implemented program simulates the process of heat transfer along the entire length of the rod of a given size using an explicit [finite-difference scheme](https://en.wikipedia.org/wiki/Finite_difference_method).  

The longer the rod (array size) - the more calculations need to be done to achieve the result.

## Requirements

* Linux machine

* [CMake 3.1 or later](https://cmake.org/download)

* [CUDA Toolkit 9](https://developer.nvidia.com/cuda-90-download-archive)

## Instructions

Compilation:

```bash

nvcc main.cu

```

Running:

```bash

[optirun] ./a.out [ARRAY SIZE] [CUDA THREADS NUMBER]

```

## Performance

CPU: Core i7-6500U CPU @ 2.50GHz ×4  

GPU: GeForce 940M

```

gleb@home: optirun ./a.out 10

>>> CPU time: 0.003 ms

>>> GPU time: 0.301 ms

>>> Rate: 0.010

gleb@home: optirun ./a.out 1000

>>> CPU time: 0.282 ms

>>> GPU time: 0.284 ms

>>> Rate: 0.992

gleb@home: optirun ./a.out 10000

>>> CPU time: 3.091 ms

>>> GPU time: 0.427 ms

>>> Rate: 7.233

gleb@home: optirun ./a.out 100000

>>> CPU time: 29.232 ms

>>> GPU time: 1.904 ms

>>> Rate: 15.353

```

The final graph of the dependence of the performance gain on the size of the array:  

![](https://github.com/rugleb/cuda/blob/master/benchmark/chart.png?raw=true)

## License

This repo is published under the MIT license, see [LICENSE](https://github.com/rugleb/cuda/blob/master/LICENSE).