hazevt04

I'm developing on a Jetson Xavier (CC 7.2) which has integrated Volta cores that share CPU memory (no cudaMemcpy() needed!) I got the idea here: https://docs.nvidia.com/cuda/cuda-for-tegra-appnote/index.html I've been following that...

I don't expect the allocator to change the attachment of the memory it allocates during the lifetime of a std::vector instance. I assume that once I set the flag for...

This isn't really an issue per se, but a gotcha I noticed. When I launched a CUDA kernel using a vector.data() pointer for output and the kernel succeeded, vector.size() was...