> The question is: Is there anything either in Linux or the 21264 that
> could account for such behavior? Could the cache be polluted by other
> processes that effectively? (The machines were basically idle during
> benchmarks).
We've seen similar behavior on linux 2.3.99 using lmbench. If you're
using linux 2.3, you may want to look at Joe Martin's page coloring
which was posted to the linux list a while back. When we tried this
it seemed to improve things (i.e. lmbench was able to correctly size
the L2 cache in its memory latency test).
However there was some concern about this page coloring patch due to the
amount of system overhead it introduced.
FWIW,
Pat
--
Patrick O'Rourke
[EMAIL PROTECTED]
