NUMA nodes usually represents two or more pieces of silicon, each with memory connected to them directly, but where both are also interconnected into a proper SMP system using some bus like DMI.
Which means memory all memory is logically accessible to all processors same ways as UMA SMP system, but accesses to local memory inside a NUMA domain has lower latency due both physics and avoiding DMI overhead. With all that, I don't think GRUB cares that much about memory latency :) -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2106208 Title: Relocation overflow in GRUB when booting with RISC-V EDK II and emulating NUMA. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/edk2/+bug/2106208/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
