On Tue, Jan 30, 2024 at 10:16:13AM +0000, Xiong Yining wrote: > Hi, Russell > > > > > Needless to say, the performance results from kernel text replication > > are workload specific, but appear to show a gain of between 6% and > > 17% for database-centric like workloads. When combined with userspace > > awareness of NUMA, this can result in a gain of over 50%. > > > > Tested-off-by: Xiong Yining <xiongyining1480@xxxxxxxxxxxxxx> > Tested-off-by: Yuquan Wang <wangyuquan1236@xxxxxxxxxxxxxx> > > Sorry for disturbing. This time I have applyed this patch on a real NUMA machine > through "-ktext=1" linux options and successfully booted. With a database-centric > performance test, it shows the patch will provide some improvement near 10%. However, > since I am unfamiliar with numa performace test, could you support more details or > suggestions on how to get the "gain of over 50%"? You need to profile the system, and work out where the performance bottlenecks are, and come up with ideas to address those bottlenecks. Kernel text replication does not give arbitary guarantees of improvement, so pulling figures of "gain of over 50%" is a complete misrepresentation - it's highly workload and platform specific. -- RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
