CUDA pass c array of GpuMat to kernel

Question

I'm new to CUDA and C and I could use some help with the following: I want to pass a C array of GpuMats to a CUDA kernel:

Here is the code of my Kernel:

__global__
    void disparityFromDiffMapsKernel(cuda::PtrStepSzi* differenceMapsArray,
                                 int arraySize,
                                 cuda::PtrStepSzi disparityMap){
    int x = blockIdx.x * blockDim.x + threadIdx.x;
    int y = blockIdx.y * blockDim.y + threadIdx.y;



    //check if thread is inside the image
    if(x > differenceMapsArray[0].cols || y > differenceMapsArray[0].rows){
        return;
    }

    //do stuff

}

And here is the code where I initialize the array and call the kernel:

cuda::PtrStepSzi diffMaps[diffMapsSize];
for(int i = 0; i <= offset; i++){
    cuda::GpuMat diffMap(leftImageGPU.size(),CV_32SC1);
    cuda::PtrStepSzi diffMapPtr = diffMap;
    diffMaps[i] = diffMapPtr;
}

disparityFromDiffMapsKernel<<<numBlocks,threadsPerBlock>>>(diffMaps,diffMapsSize,disparityImageGPU); //gpu mat is initialized before

When I run this code I get the following opencv error:

OpenCV(3.4.1) Error: Gpu API call (an illegal memory access was encountered)

I would really appreciate any help!

diffMaps resides on the host, while you are passing it as a kernel argument and trying to access it on the device. — sgarizvi
– sgarizvi, Commented Dec 18, 2018 at 5:54

Simon Doll · Accepted Answer · 2018-12-18 10:02:19Z

I found a solution to my problem by moving the array to the gpu memory via cudaMalloc and cudaMemcpy (thanks to @sgarizvi for the hint)

Here is the final code, in case someone has a similar problem:

// reserve memory for the diffmap ptrs arrays
cuda::PtrStepSzi* cpuDiffMapPtrs;
cpuDiffMapPtrs = (cuda::PtrStepSzi*) malloc(diffMapsSize * sizeof(cuda::PtrStepSzi));

cuda::PtrStepSzi* gpuDiffMapPtrs;
cudaMalloc(&gpuDiffMapPtrs, diffMapsSize * sizeof(cuda::PtrStepSzi));

//fill cpu array with ptrs to gpu mats
for(int i = 0; i< diffMapsSize; i++){
    cuda::GpuMat diffMap(leftImageGPU.size(),CV_32SC1);
    //do stuff with the gpu mats
    cpuDiffMapPtrs[i] = diffMap;
}

//copy cpu array to gpu
cudaMemcpy(gpuDiffMapPtrs,cpuDiffMapPtrs,diffMapsSize * sizeof(cuda::PtrStepSzi), cudaMemcpyHostToDevice);



disparityFromDiffMapsKernel<<<numBlocks,threadsPerBlock>>>(gpuDiffMapPtrs,diffMapsSize,halfKernelSize,disparityImageGPU);

// free the allocated memory
cudaFree(gpuDiffMapPtrs);
free(cpuDiffMapPtrs);

Collectives™ on Stack Overflow

CUDA pass c array of GpuMat to kernel

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related