Some explanation on the above extreme difference. In the misc rw benchmark the
reason in the same amount of time I get a total number of down 3332966 and you
get only 819163 is that I provide recursive down_read and that in turn can
starve the down_write (my first patches weren't implementing fair semaphores).
As you can see in my post of yesterday I made my semaphores fair in my last
patches (from rwsem-generic-5). (you didn't said which patch you used exactly
but obviously it was earlier than the -5 revision)
I'm uncertain if I should drop the list_empty() check from the fast path and if
I should still allow up_* to be called from irq/softirq, if I reduce the max
number of sleepers to 2^16 and I will provide weaker wakeup semantics I won't
be penalizied anymore and then we'll really compare apples to orange making the
comparison more interesting (probably I will do because later on I can probably
re-add that two features without too much pain).
About the benchmark you wrote it looks good measure to me, thanks.
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/