[let_me_post_0]

Nvidia isn't very big on opensource either. Most CUDA libraries are still closed source. I think this might eventually be their downfall, because people want to know what they are working with. For example with PyTorch, I can profile the library against my use case and then decide to modify the official library to get some bespoke optimization. With CUDA, if I need to do that, I need to start from scratch and guess as to whether the library from the api already has such optimizations.

[einpoklum]

NVIDIA does have a bunch of FOSS libraries - like CUB and Thrust (now part of CCCL). But - they tend to suffer from "not invented here" syndrome [1] ; so they seem to avoid collaboration on FOSS they don't manage/control by themselves.

I have a bit of a chip on my shoulder here, since I've been trying to pitch my Modern C++ API wrappers to them for years, and even though I've gotten some appreciative comments from individuals, they have shown zero interest.

https://github.com/eyalroz/cuda-api-wrappers/

There is also their driver, which is supposedly "open source", but actually none of the logic is exposed to you. Their runtime library is closed too, their management utility (nvidia-smi), their LLVM-based compiler, their profilers, their OpenCL stack :-(

I must say they do have relatively extensive documentation, even if it doesn't cover everything.

[1] - https://en.wikipedia.org/wiki/Not_invented_here

[TerraHertz]

There's a deeper reason. Remember 3Dfx? They made the entire source code for their 3D hardware available to developers, all in C and a tiny bit of assembler. It could be easily ported to non-Wintel platforms. (I know, because I did port it to a MIPS based platform that had zero operating system. It was a poker machine.)

Then 3Dfx was smashed from the inside and bought out by nVidia. Source code to 3D accellerator hardware drivers never to be seen again.

Why? Because if just anybody could port 3D graphics hardware and drivers to any custom hardware and OS platform, then Microsoft, Apple, etc would rapidly be killed by something with a MUCH better GUI (3D) appearing on the market.

The powers that be do NOT want capable, unchained computing systems to upset their carefully constructed 'enslavement via enshitification' schemes.

[ryao]

Some of it is open source:

https://github.com/NVIDIA/open-gpu-kernel-modules

[einpoklum]

Have you actually looked at those "sources"? You should. Or rather, you shouldn't since you'd be wasting your time trying to find anything useful.

[ryao]

I have peeked at parts. I enjoyed seeing that they use coverity on the modules.

[garaetjjte]

>Source code to 3D accellerator hardware drivers never to be seen again.

What? Mesa supports plenty of hardware.