Setup your machine

Steps to setup your machine in order to be able to work with cvGPUSpeedup and FusedKernel

System minimum requirements

NVIDIA

You need a PC with an NVIDIA GPU of any architecture supported by nvcc 11.8 and nvcc 12.3.

AMD

We expect the library to be compilable with ROCm HIP, but we did not test it. Check ROCm AMD hardware compatibility to know which AMD GPU's you could test. Actually, if you try with an AMD GPU we would love you to let us know :)

Intel

Intel GPU's are programmed with either OpenCL or SYCL. At this moment we don't support neither of the two API's.  

Common steps

Linux

• CMake 3.22 or higher.
• CUDA Toolkit SDK tested versions (11.8, 12.1, 12.2, 12.3) compatible with your OS.

  DO NOT install versions 12.4 or higher, since their nvcc compiler has a bug that is still not fixed.

• NVIDIA driver version required by the SDK or higher.

Windows

Install the minimum software requirements:

• CMake 3.22 or higher.
• Visual Studio 2022 Community version 17.9.5 or lower (versions 17.10 and higher require newer versions of the cuda compiler)
• CUDA Toolkit SDK tested versions (11.8, 12.1, 12.2, 12.3)

  DO NOT install versions 12.4 or higher, since their nvcc compiler has a bug that is still not fixed.

• NVIDIA driver version required by the SDK or higher.

To use FusedKernel in your code

Linux

Windows

To use cvGPUSpeedup in your code

Linux

Windows

To compile and execute the Fused Kernel unit tests

Linux

Windows

To compile and execute the cvGPUSpeedup unit tests

Linux

Windows

To generate the doxygen documentation

Linux

Windows