I would say it's worth backporting given it's a 6-18% performance penalty not backporting the patch? At least in my lxd podman workloads.
1) In the Pi 5 bootloader firmware (specifically for the BCM2712 SoC), Raspberry Pi engineers introduced a feature called SDRAM_BANKLOW in the EEPROM to reconfigure the address bit mapping across physical LPDDR4X SDRAM banks. When SDRAM_BANKLOW is active, the Raspberry Pi bootloader automatically injects numa_policy=interleave and numa=fake=8 into the kernel command line. This partitions memory into 8 virtual nodes so the Linux kernel round-robins page allocations across different physical SDRAM banks. https://forums.raspberrypi.com/viewtopic.php?t=378276 https://www.jeffgeerling.com/blog/2024/numa-emulation-speeds-pi-5-and- other-improvements/ > There were observations that 8GB Pi's were (for specific tests) slower than 4GB. Given my pi5 8gb it would be nice if this patch was backported and implemented. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2166421 Title: Malformed early option 'numa' passed in numa=fake=8 To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-raspi/+bug/2166421/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
