I think that my simulation was bogus. This because I guess that
return_all will invalidate any kind of return prediction... So the
patch could improve things even if the simulation didn' t
agreed. Is there any volunteers that could do a benchmark of
2.1.130 and 2.1.130 + my asm patch? To benchmark you should
generate a signal flood to run many times a signal handler or
something similar... I am not going to benchmark myself since my
hardware seems to not care too much about the return prediction ;)
I believe lat_sig from lmbench would be a good way to generate such a
sequence of events.
Later,
David S. Miller
davem@dm.cobaltmicro.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/