Getting Started


I am new to GPU and CUDA. I am looking into buying a TESLA product and want to do some monte carlo simulations on it. I know there is a sample program.

My question though is how do I get started. I need the hardware and need to plug it in. I need to download the CUDA libraries. What next??? Can I use VIsual Studio C/C++???

I am familiar with C. I am looking for step by step instruction on writing my first program and some introductory material about GPUs and their advantages.



you’ll need a driver with CUDA 1.1 suport, the Toolkit (cuda compiler, ptx assembler, manuals) and the SDK with a lot of demo projects.

You get it all at:

NV Developer GPU Computing Page

Look at the SDK samples and the CUDA Programming Guide.

The windows version of the SDK uses VS2003/2005 projects. The Cuda files (source code files with .cu) are splitted into host (CPU) and device (GPU) code and the host code is compiled by VC++. The GPU code is compiled by nvcc to PTX (data parallel virtual machine) assembly code and this is assembled to a cubin file with native G80 code (and I think opcodes for the PTX VM, so this could be optimized for later CUDA architectures - so it’s a kind of “fat binary”)

There is also a course from University of Illinois about CUDA with slides and MP3s:

Course University Illinois

Have fun!