VulcanicCythonNet - Devember 2020

So.

I have always really liked Level1tech. But I have not really been apart of the community yet. So maybe devember is the thing that will introduce me to you guys in a good way.

This project I have been thinking about for a while. I have been studying Data Science for about 2 years now. This semester I have had a course in OS’s and C.

People usually say that python is slow. I want to do a project were I show how it can be done. That speed can be achieved with a combination of Python, Cython and Vulkan.

Why Vulkan? Because I spend way to long trying to make my GTX 770 work with Tensorflow 2.0 one year ago. Cuda is not where I want to go.

I will build a python extension to networkx that will calculate betweness centrality of a network with vulkan running on a GPU. I will do this with Cython to be able to integrate it into any python program. I will be happy if I get it to work. I will be very grateful if my calculations are faster than native networkx.

Wish me luck.

I have heard that Vulkan is hard.

4 Likes

Good luck!
Vulkan isn’t so bad its a lot of reading of the docs, dotting your i’s and crossing your t’s. The first thing you should do when you start working with vulkan is to enable the validation layers. They will tell you if you are using the API in an invalid way.

  • Somebody that made a ray tracer using vulkan compute shaders.
2 Likes