I can create a VBO as a linear float array consisting of x,y,z values. I wonder if it is possible to:
- Use the array pointer returned by cudaGLMapBufferObject as a parameter in CUBLAS function.
- If yes, with which function I multiply all x,y,z values with a predefined 4*4 matrix with CUBlas.
- If no, is it appropriate to write a kernel function to make the mentioned multiplication. What would be the optimal grid, block etc. size to execute the multiplication kernel on a linear array for matrix multiplication.