[Qemu-devel] Benchmarks reliability

[Patrizio Boschi]

I was doing some benchmarks (with lmbench, dbench, whetstone and
compile times) to figure out guest/host performance ratios.
Are the results reliable? When benching the guest, i'm actually using
a  little client/server program to launch some workloads on the guest
and have them timed (with gettimeofday) both on the guest and the
host; for now i always got the same values (except when i manually
stop and resume the guest), e.g. 12345 msecs on the guest are also
~12345 msecs on the host. But i'd like to know if this is always true,
because i can't use this approach for some benchmarks.

Thanks,
Patrizio