Editing Vulkan Image in CUDA - external memory and interop problem

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
I have read multiple examples of how to do this such as this, this and this. There are, however, still issues that I cannot solve. Any help would be appreciated.

I have created this function to get the handle for the Vulkan Image memory:

int GpuVulkan::getResultHandle()
{
	vk::MemoryGetFdInfoKHR memoryGetInfo;
	memoryGetInfo   .setPNext(nullptr)
					.setMemory(textures.images[0].image.textureImageMemory.get())
					.setHandleType(vk::ExternalMemoryHandleTypeFlagBitsKHR::eOpaqueFd);
	int handle{-1};
	auto getMemoryInfoFdKHR = (PFN_vkGetMemoryFdKHR)device->getProcAddr("vkGetMemoryFdKHR");
	auto mgi = static_cast<VkMemoryGetFdInfoKHR>(memoryGetInfo);
	getMemoryInfoFdKHR(device.get(), &mgi ,&handle);
	return handle;
}

Then I have a class for the CUDA computations where I wanted to map this image and write something in it:

ComputeCuda::ComputeCuda(int rh, int width, int height) : Compute(rh)
{
    CUDA_EXTERNAL_MEMORY_HANDLE_DESC handleDesc;
    handleDesc.type = CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD;
    handleDesc.handle.fd = rh;
    handleDesc.size = width*height*4;

    cuImportExternalMemory(&externalImageMemory, &handleDesc);
    
    CUmipmappedArray mipmapArray;

    CUDA_ARRAY3D_DESCRIPTOR arrayDesc = { };
    arrayDesc.Width = width;
    arrayDesc.Height = height;
    arrayDesc.Depth = 0;
    arrayDesc.Format = CU_AD_FORMAT_UNSIGNED_INT8;
    arrayDesc.NumChannels = 4;
    arrayDesc.Flags = CUDA_ARRAY3D_COLOR_ATTACHMENT;

    CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC mipmapArrayDesc;
    mipmapArrayDesc.arrayDesc = arrayDesc;
    mipmapArrayDesc.numLevels = 1;
    mipmapArrayDesc.offset = 0;
    
    cuExternalMemoryGetMappedMipmappedArray(&mipmapArray, externalImageMemory, &mipmapArrayDesc);
    std::cerr << "error:" << cuMipmappedArrayGetLevel(reinterpret_cast<CUarray*>(&resultArray), mipmapArray, 0);
}

__global__ void test(cudaSurfaceObject_t result)
{
    unsigned int x = blockIdx.x * blockDim.x + threadIdx.x;
    unsigned int y = blockIdx.y * blockDim.y + threadIdx.y;
    uchar4 c = {255,255,255,255};
    surf2Dwrite(c ,result, x*sizeof(uchar3), y);
}

void ComputeCuda::compute()
{
    cudaResourceDesc rd;
    memset(&rd, 0, sizeof(rd));
    rd.resType = cudaResourceTypeArray;
    rd.res.array.array = resultArray;
    std::cerr <<"aaa";
    cudaCreateSurfaceObject(&resultSurface, &rd);
    test<<<10,256>>>(resultSurface);  
}

I have struggled with some deprecated examples of the external memory management and also the type casting such as the reinterpret_cast that I had to use and didn’t see it in another examples but I think that this should be working well. However, I got a Segmentation Fault at the cudaCreateSurfaceObject function and 709 cudaErrorContextIsDestroyed error at cuMipmappedArrayGetLevel. I am now kinda clueless about where the problem might be.
Thanks and have a nice day

EDIT:
Had to use cudaSetDevice and cuCtxCreate to fix the context error but still getting Bus error instead.