parallel.gpu.CUDAKernel on MATLAB Home Edition

So, the idea of MATLAB Home, I assume, is to provide an inexpensive option for home users. However, if I want to use parallel.gpu.CUDAKernel objects from the Parallel Computing Toolbox, I need some way of compiling .cu files to .ptx and, under Windows at least, the NVIDIA CUDA Toolkit only supports Visual Studio Professional. Are there any cheap workarounds for doing ptx compilation, as a home user, by which I could avoid shelling out the bucks for VS Pro?

Answer 1

[Edric Ellis]

You could try this workaround: http://stackoverflow.com/questions/8900617/how-can-i-setup-nvcc-to-use-visual-c-express-2010-x64-from-windows-sdk-7-1. If it's an option for you, the toolchain to build PTX files is available for free on Linux ( nvcc + gcc ). (You can usually build your PTX on a 64-bit Linux system and then run it on a 64-bit Windows system).

[Matt J]

Thanks, Edric. I'll give it a try.

I was wondering, though. Would it be prohibitively difficult for the Parallel Computing Toolbox to include its own compiling commands for CUDA code, similar to mex()?