Deploying Application on Windows targetting different GPU

Software
1

Hi there,

I have compiled the spatial mapping example on Windows using this guide and would like to deploy the resulting .exe onto another Windows PC

My First PC has a NVIDIA RTX A2000 Laptop GPU, and GPU driver version (528.90) , with CUDA 11.8 toolkit installed.

I know CUDA 11.X is compatible with a minimum Nvidia Compute Capability of 3.5, but when compiling the spatial mapping example, I am unsure as to which GPU architectures the ZED application is compiling for…

  • When compiling the ZED SDK samples, which architectures of NVidia GPUs are targeted? Is it currently to set to only compile for the currently discovered GPU?
  • Am I able to compile for all GPU versions that are compatible with CUDA 11.X? e.g. Will the exe compiled on my laptop with an RTX A2000 (compute capability 8.6) be able to be used with a PC with a GTX 1080 (compute capability: 6)
  • If not, how can I target all compatible GPU versions (say, compute capability > 6.0) ?
2

Indeed if you build your application it will be compatible with all machine that are running the CUDA version that the builder has.
It’s the same for SDK samples. It’s not about the GPU itself, it’s about the version of CUDA.