Hello,
I am currently exploring the use of MathNet.Numerics.Control.UseNativeCUDA() for enabling GPU acceleration in my application (.Net6), but I have encountered an issue that I have been unable to resolve.
My setup includes:
•GPU: NVIDIA GeForce GTX 1650
•CUDA Toolkit Version: 12.8
•Installed Package: MathNet.Numerics.Providers.CUDA (from NuGet)
When attempting to enable CUDA with MathNet.Numerics.Control.UseNativeCUDA(), I receive the following error:
System.NotSupportedException: "Cuda Native Provider does not support capability querying and is therefore not compatible. Consider upgrading to a newer version."
EntryPointNotFoundException: Unable to find an entry point named 'query_capability' in DLL 'libMathNetNumericsCUDA'.
From my research, I suspect that this issue may stem from a mismatch between the library and the GPU or CUDA version. However, I have verified that the GeForce GTX 1650 (compute capability 7.5) should support CUDA, and other GPU computing frameworks like ILGPU are functioning as expected on my system.
I would greatly appreciate your guidance on the following questions: - What is the recommended CUDA version for MathNet.Numerics.Providers.CUDA?
- Is there a known compatibility issue with the GeForce GTX 1650 or similar GPUs?
- Are there specific steps or configurations required to resolve this exception?
Thank you for your time and assistance.
|