is anyone using HZ=1000 plus kernel profiling? The profiling output
should get sqrt(10) more precision this way ...
[just wanted to know wether i have to back up my HD before trying
it :)]
It has been done on the Sparc. In fact I think we got it clicking up
to a HZ of 10,000 at one point. The profiling was so fine that you
could actually see exactly which load/store instructions were missing
the cache more often than others in memcpy/memset/csum_partial etc.
Pipeline stalls were extremely visible as well.
---------------------------------------------////
Yow! 11.26 MB/s remote host TCP bandwidth & ////
199 usec remote TCP latency over 100Mb/s ////
ethernet. Beat that! ////
-----------------------------------------////__________ o
David S. Miller, davem@caip.rutgers.edu /_____________/ / // /_/ ><