DRIVE OS Version: 6.0.10.0-36101120
Hardware: Nvidia Drive AGX Orin (p3710)
Yocto linux: 3.1 (dunfell)
Issue Description:
I have a similar issue to this thread: How to get Yocto cuda recpies to package x86 nvcc in sdk?
I would like to include the CUDA components to the host sysroot of the Yocto generated SDK to be able to cross-compile applications to the target image. However, the existing Yocto cuda recipes either generating packages to the target SDK or for native use which is used by bitbake during build (not as “native-sdk” package)
Do you have any suggestion how could I build the CUDA components as a native-sdk package in Yocto?
Thank you in advance!
Dear @roland.sagi ,
Does that mean you want to croos compile CUDA sample for yocto on docker ?
Hi @SivaRamaKrishnaNV ,
Not exactly. Right now I’m generating an SDK using the Yocto environment. This SDK can be used to cross compile applications to Orin.
However, for the cross compilation I need the CUDA components (not just the sample app, the entire “package”) to be part of the host sysroot of this SDK, since the cross compilation process will be executed in the x86 docker environment (I think this is what you were referring to).
My question is, do you have any suggestion on how should I modify the recipes so that the CUDA components will be included in the host sysroot of the Yocto generated SDK.
Thank you,
Roland
Could you please provide any update for this topic?
Does this topic still need support? It is a blocker for your application development or just trying to evaluate to see if it works? I don’t see steps in documentation to support this.
Are you planning to include the CUDA components in a future release or could you provide a patch to modify the existing bitbake recipes to get the expected result?
It is not a blocker for us at the moment, but it would be really useful if we could use the SDK generated by the yocto environment.