During some research I analysed the spinlock code of Linux, Zephyr and Arm and discovered that the spinlock code of Xen on Arm could probably be optimized which could improve the performance. The fact that the spinlock code is done in C in the common code is not helping to optimize it cleanly. I would like to have other developer view on this and see if this is something that we could work on.