ansaurus

Question

How to integrate CUDA .cu code with C++ app

Answer 1

+2 A:

It looks like you have a conflict between two different versions of the C runtime. The CUDA runtime uses the statically linked C runtime, and your application uses the dynamically loaded C runtime (typically the default in Visual Studio).

Change the code generation to use statically loaded C runtime to match the CUDA runtime:

Right click on the project and choose Properties, then in C/C++ -> Code Generation change the Runtime Library to /MT

You could use /MTd for debug, in which case you will need to mirror this in CUDA Build Rule -> Hybrid CUDA/C++ Options.

See this earlier post for a step-by-step with more tips.

Tom 2010-01-20 11:39:04

Thanks Tom. I think you have mentioned this before on an earlier post of mine. I'll try it when I get home tonight.

Mr Bell 2010-01-20 14:48:21

Using /MTd allows everything to compile. However, I switched the runtime library to /MT and ensured that CUDA settings matched. This yielded a new set of errors, i'll post the new build log in the post.Thank you for your continuing insight

Mr Bell 2010-01-21 03:16:31

/MD and /MDd both work fine. Whats the main difference between /MT and /MD ?

Mr Bell 2010-01-21 03:40:19

Check out http://msdn.microsoft.com/en-us/library/2kzt1wy3.aspx for a description of the difference between /MD or /MT - too long for a comment!

Tom 2010-01-21 08:44:25

Regarding your latest error message, it looks like you have set /MT in the CUDA Build Rule but not in the C/C++ Code Generation page. Double check that you have made the change in *both* places, and then rebuild everything.

Tom 2010-01-21 08:46:44

Regarding my above comment, I think you actually have /MT in the CUDA build rule and /MTd in the C/C++ settings. They need to match, either both /MT or both /MTd.

Tom 2010-01-21 10:08:18

Answer 2

A:

I am a newbie to CUDA and visual studio!! Mr.Bell , it seems that u have been able to compile CUDA codes in VS 2008. I m running VS 2008 on WinVista. I have integrated CUDA sdk with VS 2008.

1) I need to know that CUDA codes are to be compiled as .cpp or .cu, saving the files as .cu and then compiling is giving a lot of errors. As I am a newbie I do not know how to proceed further. Ur advice would be of gr8 help. I am compiling a simple Hello world code.

I m pasting the code for ur reference

* ** Hello World using CUDA ** ** The string "Hello World!" is mangled then restored using a common CUDA idiom ** ** Byron Galbraith ** 2009-02-18 */

include

// Prototypes global void helloWorld(char*);

// Host function int main(int argc, char** argv) { int i;

// desired output char str[] = "Hello World!";

// mangle contents of output // the null character is left intact for simplicity for(i = 0; i < 12; i++) str[i] -= i;

// allocate memory on the device char *d_str; size_t size = sizeof(str); cudaMalloc((void**)&d_str, size);

// copy the string to the device cudaMemcpy(d_str, str, size, cudaMemcpyHostToDevice);

// set the grid and block sizes dim3 dimGrid(2); // one block per word
dim3 dimBlock(6); // one thread per character

// invoke the kernel helloWorld<<< dimGrid, dimBlock >>>(d_str);

// retrieve the results from the device cudaMemcpy(str, d_str, size, cudaMemcpyDeviceToHost);

// free up the allocated memory on the device cudaFree(d_str);

// everyone's favorite part printf("%s\n", str);

return 0; }

// Device kernel global void helloWorld(char* str) { // determine where in the thread grid we are int idx = blockIdx.x * blockDim.x + threadIdx.x;

// unmangle output str[idx] += idx; }

ankit 2010-06-16 13:22:24

ansaurus

tags:

views:

answers:

How to integrate CUDA .cu code with C++ app

Edit:

include

include

related questions