Not really. There is minimal support for applications that want to inquire about their NUMA configuration and learn that there is just one NUMA node. But there is no support for features like:
- Retrieving from the hypervisor (emulated hardware) information about the NUMA configuration and making this information available to the kernel and applications.
- Allowing applications to allocate memory from the nearest NUMA node.
It will be great if someone who is familiar with libnuma and what's missing would open issues on what exactly is missing, and what are the reasonable priorities for implementing them (i.e., which features are really important in modern VMs?).