CUDA large vectors cause errors

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m trying to copy some memory over to the device, but I’m having some trouble with larger vectors, the following vectors works perfectly:

thrust::device_vector<unsigned int> data = std::vector<unsigned int>(115200); // smaller size

But this one does not:

thrust::device_vector<unsigned int> data = std::vector<unsigned int>(1152000); // (roughly 5 MB worth of data)

When I get a pointer from the larger vector (thrust::raw_pointer_cast) and use it in a kernel like this I get an access violation error after the kernel finishes running, the smaller vector works just fine.

__global__ void Kernel(unsigned int* data) {
    printf("%u\n", data[0]);
}

Is there perhaps a size limit on device vectors, or is the size type overflowing?
Any advice is greatly appreciated!

2

The only size limited on vector will be the available global memory available on device.
It’s much easier to help you if you provide an entire sample.

3

Hey, I’ve done some dug around in my code base, and found out that the error is being caused by something a bit different, I’ve also created a minimal viable example to showcase the behavior of my code. The new question can be found here. Regards