TextureConvolution using CUDA or OpenCL

Demonstates how to use the InventorCuda or InventorCL API and the InvCudaAlgorithms in order to achieve high performence filtering using the CUDA/OpenCL API.

This demo can perform Gaussian Blur, sobel or Prewitt filtering using the regular cpu, the CUDA or the OpenCL based convolution algorithms. It is also possible to specify a custom filtering matrix. Only separable matrices are supported at this time. When the compute mode is switched from the CPU mode to the CUDA or OpenCL mode the Open Inventor module is loaded using dynamic module loading mechanism and the new algorithms interface is automaticaly used. The SoAlgorithms class brings the mechanism for this automatic usage of the loaded API.

FILES:

• Inventor/examples/Features/Compute/TextureConvolution/main.cxx
• Inventor/examples/Features/Compute/TextureConvolution/Filtering.cpp
• Inventor/examples/Features/Compute/TextureConvolution/Filtering.h
• Inventor/examples/Features/Compute/TextureConvolution/Data/gui.iv

SEE ALSO

SoCudaApi, SoOpenCLApi, SoAlgorithms

SCREENSHOT: