On Thu, Mar 12, 2015 at 12:30 PM, Ian Campbell <ian.campb...@citrix.com> wrote:
> On Fri, 2015-03-06 at 22:24 +0100, Tamas K Lengyel wrote: > > The ARM virtualization extension provides 2-stage paging, a similar > mechanisms > > to Intel's EPT, which can be used to trace the memory accesses performed > by > > the guest systems. This series sets up the necessary infrastructure in > the > > ARM code to deliver the event on R/W/X traps. Finally, we turn on the > > compilation of mem_access and mem_event on ARM and perform the necessary > > changes to the tools side. > > > > This version of the series is based on master and each patch in the > series has > > been compile tested on both ARM and x86. > > > > The following benchmarks have been completed on an Arndale board before > and > > after applying this entire series in dom0, which had 2 vCPUs available > > pinned. No regressions are evident based on these benchmarks. > > > > | f0ffd60 | AFTER > > -------------------------------------------- > > hackbench | 1.604 | 1.618 > > hackbench -pT | 1.170 | 1.126 > > hardinfo CPU Blowfish | 43.252 | 43.323 > > hardinfo CPU CryptoHash | 40.935 | 40.893 > > hardinfo CPU Fibonacci | 10.358 | 10.360 > > hardinfo CPU N-Queens | 25.773 | 25.894 > > hardinfo FPU FFT | 20.332 | 20.325 > > hardinfo FPU Raytracing | 73.017 | 73.104 > > Thanks for these, they are very reassuring. > > More for curiosities sake than anything else: Do you have any figures > for the overhead of enabling memaccess? > I can gather some figures if needed but it is very use-case specific depending on which pages are being trapped and how. Tamas
_______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
