Seongbae Park wrote:
membar #storestore is effectively a nop on all existing sparc processors because they only support TSO mode. So there's no performance penalty.
On SPARC, yes. Not so on x86 or amd64 unfortunately with sfence.
Is there any chance that cpu_mstate_gen wraps around between two loads of cpu_mstate_gen in get_cpu_mstate(), making them return the same value ? Not very likely but is there anything that will prevent such a case ?
It is possible, if incredibly unlikely. Such an event seems to imply a big stall on the kstat reader's side, in which case we've got other problems and the reader's results won't be consistent no matter what we do. I don't think I can construct a scenario here that will lead to any large kstat error, and any small error will become corrected after a little time.
-- Ethan _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
