Konstantin: > write locks can be disregarded here. its only read locks which cannot > be executed in parallel. nobody blames os and/or hardware. > Im trying to understand is it expected behavior or not.
I'm not sure I understand why you're saying that it's okay to disregard write locks. Are you able to explain what kind of data structure this RW lock is protecting? > what I see is that 10 or so threads read locking same RW lock cannot > execute simultaneously. this happens only on hardware where this > threads can execute simultaneously and not serialized due to lower > number of cpus. Perhaps you're suffering from the effects of false sharing? This might explain why you see the problem on a 16 socket box, but not on a T2000. If you are, the effects probably won't be as bad on a box that behaves more like a traditional SMP system. Is this RW lock embedded within a data structure, statically declared somewhere, or something entirely different? -j _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
