On 8 November 2012 14:53, Sundaram, Senthilkumar <ssundara@xxxxxxxxxxxxxxxx> wrote: > Peter Maydell [mailto:peter.maydell@xxxxxxxxxx] wrote: >> You should be aware that the performance characteristics of the fast model >> do not particularly correspond to how any kind of hardware might behave; >> you need to be very cautious about interpreting cpu utilization or other >> performance related data you gather. Gathering perf info on real hardware >> would be rather safer. >> > [[ss]] Agreed. I am waiting for my hardware. I want to have my setup > ready by the time the hardware shows up. Also, while the timing > information may not be very useful, the "kvm trace events" may still > be useful. Perf has a mode to print the count of various events that > occurred for a particular period. I think that will be still be useful. Yes, that all sounds reasonable -- just wanted to check you were aware of the potential pitfalls in this area. -- PMM _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
