Hello, I am a student studying algorithm optimization using CUDA programming.
Is there a way to use tensor cores on Windows 10 and Visual Studio 2019?
My current environment is as follows:
GPU: RTX2080Ti, RTX3060, RTX3090
CPU: AMD RYZEN 9 5900X 12-Core Processor
Visual Studio 2019
Language: C
Thank you
Sure, there are many examples of it. If you do a CUBLAS gemm operation with appropriate API choices, it will use the GPU tensor cores. It’s not OS-specific.
CUDA library examples are available here. As an example, one code that is designed to use tensor cores is the LtIgemm sample. Here is another example showing both INT8 and FP16 usage of tensorcores. With a bit of googling you can find many more examples.
Thank you for your replying.
It helped me a lot