Il 15/05/2013 18:16, Peter Maydell ha scritto: > On 15 May 2013 16:48, Paolo Bonzini <pbonz...@redhat.com> wrote: >> Giving a shot to rcutorture on a weak memory-model machine (PPC) >> would be nice. > > # ./tests/rcutorture 16 rperf 2 > n_reads: 116103000 n_updates: 0 nreaders: 16 nupdaters: 0 duration: 2 > ns/read: 275.617 ns/update: nan > > ...that "nan" isn't entirely reassuring :-) > > Other than that, I gave it a few runs and it seemed OK, but it > would be helpful if you could give some command lines that would > constitute a usefully thorough test process.
"tests/rcutorture 10 stress 10" is a pretty good test. Here it gives something like this: n_reads: 30374101 n_updates: 232950 n_mberror: 0 rcu_stress_count: 30028631 345470 0 0 0 0 0 0 0 0 0 It will abort if something goes wrong. Paolo
