OptiX 8.1 Code Differences

[devillove084]

@LegNeato I've installed OptiX 8.1. After generating the code, there are significant differences compared to the current code. I'm wondering if it's still necessary to maintain a ray tracing sub project within the main CUDA project. Alternatively, should I consider moving it to a dedicated "rust-optix" project so that we can focus more on the relevant issues?

[LegNeato]

I haven't dug into this area yet, so I have no opinion 😄 . If others do, please chime in!

[trigpolynom]

I ran into this when generating Optix v 9.0. I was going to start doing what @devillove084 suggested with a rust-optix crate but for version 9.

[jorge-ortega]

We're close to getting Optix 7.3 working again with the backend. It's not 8 or 9, but it's something :^)

My thoughts on spinning off the optix crates are mixed. Some things to consider:

• Stability of the NVVM backend and cust/cuda crates.
• Whether the spinoff will support 7 or focus on 8/9.

In the long term, it makes sense to create a separate project with its own CI to support multiple versions of OptiX, release independently, and let users choose versions based on their compute targets. For now, adding support for versions 8/9 in the existing project seems worthwhile, as it allows a shared CI to resolve issues quickly.

Supporting multiple versions will be challenging. Currently, we use feature flags and cfg_if! for different point releases of 7, but this approach may not scale well for major versions. A more elegant solution likely exists.

I'd recommend waiting for at least one more dot release before deciding to spin off. We can attempt adding support for 8/9 (or even consider dropping 7) in the current repo and reassess after the next stable release. That said, if there's strong consensus to spin off now, I'm open to it.

[LegNeato]

Yeah, I agree. My preference would be keep the focus here / not split the very small and nascent community. If it becomes a problem or we think we are better off splitting down the road we'll deal with it then.