Hi, I’m having trouble to understand the behavior of cudaHostAlloc. When I allocate zero-copy memory with cudaHostAlloc with cudaHostAllocMapped and cudaHostAllocPortable I can see an increase in host memory, but there is also an increase of around 100MB in GPU memory. This happens no matter how much I allocate. For example, I tried to allocate few GB and there was still an increase of around 100MB in GPU memory.
I was looking for explanation in the documentation but I couldn’t find one. Your help would be much appreciated!