Using 64 bit version of cuda driver api.

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hello,

Cuda Driver API documentation and header is basically missing one/two things:

  1. It does not explain how to switch between 32 bit and 64 bit version of cuda driver api.

NVCUDA.DLL is mentioned no where.

I am guessing NVCUDA64.DLL is the 64 bit version ? However it’s not directly in system32 folder but somewhere else ? Locations of these DLL’s in windows should be mentioned as well just in case.

Also is nvcuda.dll just a stub ? Will it actually switch to nvcuda32.dll or nvcuda64.dll based on some kind of detection ???

Bye,
Skybuck.

#2

The switch between 32-bit and 64-bit is handled by nvcc when specification of the -m switch is given.

http://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html#options-for-altering-compiler-linker-behavior

If you link against -lcuda it will pick up the correct version of the DLL.

#3

This is not good enough information for me. I am not using C/C++ with the CUDA Driver API, instead I am using a different programming language called: Delphi.

Thus I need more “binary based” information, then C/C++ compiler/linker switch magic.