What is the difference between CUDA takeit and CUDA sdk

Question

I install CUDA on Ubuntu 14.04 and have a Maxwell card (GTX 9 ** series), and I think I installed everything correctly with the toolkit, since I can compile my samples. However, I read that in the places where I have to install the SDK (this seems to be talked about with sdk 4). I'm not sure the toolkit and sdk are different? Since I have a later card of 9 series, does this mean that CUDA 6 is working for me? Here is my version of nvcc

nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2014 NVIDIA Corporation Built on Wed_Aug_27_10:36:36_CDT_2014 Cuda compilation tools, release 6.5, V6.5.16 

I follow the book and I need to include <cutil.h> , and I cannot find this file in include, wherever I install it.

I followed this guide provided by nvidia and since I did what they say, so I got confused http://developer.download.nvidia.com/compute/cuda/6_5/rel/docs/CUDA_Getting_Started_Linux.pdf

thanks for the help

Answer 1

CUDA Toolkit is a software package that has various components. The main parts:

• CUDA SDK (compiler, NVCC, CUDA software development libraries and CUDA samples)
• GUI tools (such as Eclipse Nsight for Linux / OS X or Visual Studio Nsight for Windows)
• Nvidia driver (system driver for driving a card)

It also has many other components, such as a CUDA debugger, profiler, memory controller, etc.

The fact that you can compile and run the samples means that you probably have fully installed the Toolkit and at least have an SDK, driver, and samples.

Regarding cutil.h , doing a search in my CUDA 6.5 installation using find -L . -iname "cutil.h" find -L . -iname "cutil.h" failed. Also looking at other related questions about SO, it seems that this header file no longer exists in CUDA installations (since CUDA 5.0). However, looking at the samples, you can find several new utility headers, for example helper_cuda.h . Helpers like these should be located somewhere like /usr/local/cuda/samples/common/inc on your OS. helper_cuda.h is a header that is almost always included in my CUDA programs, as I find useful functions like checkCudaErrors() very useful.

If you follow the book, my recommendation; try compiling the code, and whenever you get an error message if the utility function is missing, search for grep in the header files included in samples/common/inc . You will most likely find the missing features of the utility, and then you can include the necessary headers accordingly.