Abstract
This paper describes the techniques used to achieve high context-switching performance on ARM processors for the L4 microkernel and a para-virtualised Linux running on top. We examine how the previously-published techniques can be used in L4 with minimal changes to the kernel API. We also propose future API changes which make it easier to maximise memory-management performance, not only on ARM but also on architectures supporting a segmented memory model.