I'm familiar with libnuma. I also know how to create, attach, and map huge page memory (using shmat, shmget, shmctl).
What I need to know is how do I get hugepage memory that's NUMA aligned to the same NUMA node as the caller's processor?
- shmat et al. don't have args for NUMA mode
- libnuma doesn't appear to have methods for shared memory
What am I missing? Or is this my misconception ... and the memory is implicitly NUMA aligned to the caller's NUMA node when attached? And I wanted pinned so it's not swapped out. As far as I know, huge page memory cannot ever be swapped.