I must admit that I don’t get the clue about this, but hopefully somebody can help:
We wanted to use CUBLAS to get the mathematics of an already existing project to be computed on the GPU (CUBLAS to get results in a first step, later on we might proceed to CUDA itself). As for now there’s none of the developers having a G8 and concerning me not even a nVidia-Card :"> so we have to develop in emu-mode.
For CUBLAS everything’s clear… linking to cublasemu.dll.
But now we need interoperability between GL and CUDA(-BLAS) to work with VBOs. That’s the point which I don’t get… Do I link to cuda.dll? But then how do I switch between emu/real-mode? Is there any way to get this done without using nvcc (as we currently write only host code)?
Thanks in advance,