Following code in os_linux.cpp in os::init_2() indicates that NUMA and adaptive sizing is incompatible with parallel gc:

    if (UseParallelGC && UseNUMA && UseLargePages && !can_commit_large_page_memory()) {
      // With SHM and HugeTLBFS large pages we cannot uncommit a page, so there's no way
      // we can make the adaptive lgrp chunk resizing work. If the user specified both
      // UseNUMA and UseLargePages (or UseSHM/UseHugeTLBFS) on the command line - warn
      // and disable adaptive resizing.
      if (UseAdaptiveSizePolicy || UseAdaptiveNUMAChunkSizing) {
        warning("UseNUMA is not fully compatible with SHM/HugeTLBFS large pages, "
                "disabling adaptive resizing (-XX:-UseAdaptiveSizePolicy -XX:-UseAdaptiveNUMAChunkSizing)");
        UseAdaptiveSizePolicy = false;
        UseAdaptiveNUMAChunkSizing = false;
      }
    }


However a few lines of code later, if ForceNUMA is set, we unconditionally set UseNUMA to true.

Investigate this contradiction.