parallel processing - how to compile .cu file for dynamic parallelism with matlab?

Question

I'm trying to run a simple example of dynamic parallelism in cuda. the code of the .cu file is

__global__ void child_launch(int *data) {
   data[threadIdx.x] = data[threadIdx.x]+1;
}

__global__ void parent_launch(int *data) {
   data[threadIdx.x] = threadIdx.x;

   __syncthreads();

   if (threadIdx.x == 0) {
       child_launch<<< 1, 256 >>>(data);
       cudaDeviceSynchronize();
   }

   __syncthreads();
}

where parent_launch is the kernel I want matlab to run, and each thread of parent_launch can run a grid of blocks with the kernel child_launch (in practice, only the 0th thread should create such a grid, but that's just an example).

I tried to run it all by compiling the .cu file into a .ptx file and then executing the following commands in matlab:

   k = parallel.gpu.CUDAKernel('file_name.ptx', 'file_name.cu');
   k.GridSize = [1,256];
   k.ThreadBlockSize = [1,256];
   r1 = feval(k, data);% data is an array of ints on the gpu

the problem is that when I tried to compile the .cu file, I got the following error:

error: kernel launch from __device__ or __global__ functions requires separate compilation mode

Does anyone know how to fix it?

question from:https://stackoverflow.com/questions/65599979/how-to-compile-cu-file-for-dynamic-parallelism-with-matlab

Answer 1

Waitting for answers