# Using OpenGL instead of CUDA for machine learning

**URL:** <https://forum.devtalk.com/t/using-opengl-instead-of-cuda-for-machine-learning/8967>\
**Category:** Data Science\
**Tags:** learning, cuda, opengl, in-the-news\
**Created:** [21 April 2021 14:47 UTC](https://forum.devtalk.com/t/using-opengl-instead-of-cuda-for-machine-learning/8967 "2021-04-21T14:47:28Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![CommunityNews](https://forum.devtalk.com/user_avatar/forum.devtalk.com/communitynews/32/1113_2.png) [@CommunityNews](https://forum.devtalk.com/u/CommunityNews)\
**Post date:** [21 April 2021 14:47 UTC](https://forum.devtalk.com/t/using-opengl-instead-of-cuda-for-machine-learning/8967/1 "2021-04-21T14:47:28Z")

</div>

> ### Summary
> 
> In this project, we
> 
> 1. Added an **OpenGL backend** for MXNet/TVM - a general-purpose tensor computation framework, so that it **automatically compiles** a Python program into an OpenGL shader that runs on the GPU on a computer that **does not have CUDA**.
> 
> 2. Explored optimizations of OpenGL shader programs so that a fundamental computation task needed in machine learning - matrix multiplication - has **comparable performance with OpenCL** on the same machine.

Read in full here:

> **[OpenGL Backend for MXNet/TVM](https://phisiart.github.io/418-proj/)**
>
> A single-page template for Jekyll based blogs (like what you'd use on github pages!)

_This thread was posted by one of our members via one of our news source trackers._