Nvidia Warp: A Python framework for high performance GPU simulation and graphics

(github.com)

490 points jarmitage | 1 comments | 14 Jun 24 13:28 UTC | HN request time: 0.206s | source

Show context

eigenvalue ◴[14 Jun 24 14:23 UTC] No.40681204[source]▶

I really like how nvidia started doing more normal open source and not locking stuff behind a login to their website. It makes it so much easier now that you can just pip install all the cuda stuff for torch and other libraries without authenticating and downloading from websites and other nonsense. I guess they realized that it was dramatically reducing the engagement with their work. If it’s open source anyway then you should make it as accessible as possible.

replies(7): >>40681241 #>>40681256 #>>40681604 #>>40683495 #>>40684963 #>>40686175 #>>40688511 #

dagmx ◴[14 Jun 24 15:08 UTC] No.40681604[source]▶

>>40681204 #

This isn’t open source. It s the equivalent of headers being available to a dylib, just that they happen to be a python API.

Most of the magic is behind closed source components, and it’s posted with a fairly restrictive license.

replies(2): >>40683396 #>>40684101 #

1. boywitharupee ◴[14 Jun 24 19:24 UTC] No.40684101[source]▶

>>40681604 #

In a similar fashion, you'll see that JAX has frontend code being open-sourced, while device-related code is distributed as binaries. For example, if you're on Google's TPU, you'll see libtpu.so, and on macOS, you'll see pjrt_plugin_metal_1.x.dylib.

The main optimizations (scheduler, vectorizer, etc.) are hidden behind these shared libraries. If open-sourced, they might reveal hints about proprietary algorithms and provide clues to various hardware components, which could potentially be exploited.

↑