Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
opencl/nvidia: properly detect the PCI domain of NVIDIA GPUs
Instead of hardwiring it to 0. NVIDIA finally implemented this in CUDA 10.2. I couldn't find it in their OpenCL extenstions, but they told me it's implemented. It's the very next extension value (0x400a), it reports the right result, and following extension values (0x400b+) still report errors. Thanks to François Rue and the wonderful PlaFRIM team for the help. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
- Loading branch information