LuxCoreRender Forums

On CUDA builds, you can now choose between OpenCL and CUDA in the addon preferences.
Devices will no longer list GPUs twice.

Good. Just a question though..

Since there is now a setting that allows switching cuda/opencl does it make sense to have separate builds for opencl and cuda support?

The plan is to have only one build that runs everywhere in the future.

That is a nice addition.
How does it work when you have both AMD and NVIdia GPU i guest you select Opencl if you want both GPU to work and CUDA if you wann't use AMD for rendering.

Btw is it possible to have compilation done and saved directly inside the build. So user do no longer see compilation ?
I mean when Luxcore tell he is doing one time compilation the result of this operation is saved somewhere so why not have this resulted file pre compiled in advance inside builds ?

I hope i'm not missing a step like each PC have a different compilation file depending on the hardware available.

Sharlybg wrote: ↑Sun May 17, 2020 7:45 pm How does it work when you have both AMD and NVIdia GPU i guest you select Opencl if you want both GPU to work and CUDA if you wann't use AMD for rendering.

Good point, I have not taken this possibility into account. Currently you have to select OpenCL in this case, which will then be used for all devices.

B.Y.O.B. wrote: ↑Sun May 17, 2020 8:04 pm

Sharlybg wrote: ↑Sun May 17, 2020 7:45 pm How does it work when you have both AMD and NVIdia GPU i guest you select Opencl if you want both GPU to work and CUDA if you wann't use AMD for rendering.

Good point, I have not taken this possibility into account. Currently you have to select OpenCL in this case, which will then be used for all devices.

For the record, LuxCore is fully capable to use OpenCL and CUDA devices at the same time.

Dade wrote: ↑Sun May 17, 2020 10:52 pm For the record, LuxCore is fully capable to use OpenCL and CUDA devices at the same time.

What I'm missing a bit is some kind of unique device ID, that would allow me to detect which devices have both OpenCL and CUDA capabilities and which don't. Currently all I can do is compare names, which are not unique.
I guess it isn't strictly necessary to have this though, since it's unlikely that Nvidia, AMD and Intel will all use the same name for some device.

@B.Y.O.B. , Nice addition I think too. Now everytime I don't change the Compute settings.

Guys, are in your plans new "paper" integrations "GSoC" (Google Summer of Code)?

I have added an interactive render settings guide.
It asks a view questions and assings render settings based on the answers.
It is meant for newer users, to show which render settings are a good idea in which scene types.
Feedback and ideas welcome.