[PATCH v3 00/19] arm64 kernel text replication

[Russell King (Oracle)]

On Tue, Jan 23, 2024 at 06:35:09PM +0800, Yuquan Wang wrote:
> > 
> > After applying these patches, we tested that our business performance
> > increased by more than 5% and the NUMA node memory bandwidth was more
> > balanced.
> > 
> 
> I have successfully applied your patches on my arm64 linux. And I could 
> start it with a qemu machine(virt). However, I don't know the way to test
> the performance it brings to the kernel. Do you have some suggestions?

Please can I make one thing utterly clear... kernel text replication
in a virtual machine generally doesn't make sense unless one can
setup the virtual machine to be truly NUMA. In other words, groups
of CPUs with their local memory and remote-node memory having higher
latency.

Kernel text replication is something which solves the problem on
bare metal NUMA machines where running kernel text that is located
in a foreign node results in the CPU running slower than it would
do if the kernel text were in its local RAM.

Unless the VM is setup in exactly that way, then kernel text
replication has no place in a VM, and probably would result in
poorer performance.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!