CUDA ptxas fatal : Unresolved extern function '_Z22mwGet​GlobalThre​adIndexv'

Hi

I used gpucoder to generate the cuda file corresponding to original function. After that i built a .PTX file from my cuda file and now i want to create CUDAkernel using parallel.gpu.CUDAKernel to speedup my function. By using k = parallel.gpu.CUDAKernel('myfunction.ptx','float *, const double *','_Z15myfunction_kernel1dP15emxArray_real_T'), I am getting an error below:

Error using parallel.gpu.CUDAKernel

An error occurred during PTX compilation of <image>.

The information log was:

The error log was:

ptxas fatal : Unresolved extern function '_Z22mwGetGlobalThreadIndexv'

The CUDA error code was: CUDA_ERROR_INVALID_PTX.

How to get rid of this error? Any help will be much appreciated. Regards

Answer 1

[Linda Koletsou Soulti]

Hello Riffat,

it seems that the kernel calls into a helper function that it was not compiled with the kernel.

The following example may be helpful:

https://www.mathworks.com/help/parallel-computing/illustrating-three-approaches-to-gpu-computing-the-mandelbrot-set.html#d123e33462

Cheers,

Linda

[Riffat Tehseen Amjad]

Dear Linda,

I don't understand your answer. Can you please further elaborate?

I am already following this example.

Thanks

Riffat

[Linda Koletsou Soulti]

Hello Riffat,

in the folder of the generated code there should be a file called MWCudaDimUtility.cu. Your device code calls into a function called mwGetGlobalThreadIndex inside this file. There are a couple of ways to resolve the issue caused by this dependency. I believe the easiest would be to copy the contents of the MWCudaDimUtility.cu file inside the file of your code, before the definition of your function. Then build this file to ptx.

Cheers,

Linda

[Riffat Tehseen Amjad]

I solved the problem by copying MWCudaDimUtility.h file, found on the web, in place of MWCudaDimUtility.h file created by gpucoder and it worked. Thanks for your help Linda!

Best,

Riffat