Re: strange `time' results on dual-processor P6 systems

David S. Miller (
Fri, 31 Jan 1997 02:47:15 -0500

Date: Thu, 30 Jan 1997 16:01:55 -0800
From: Daniel Quinlan <>

The elapsed time seems correct on both systems. However, the
proportion of user to system time is more or less reversed and quite
wrong for the dual-processor system.

Any ideas?

The CPU's are spinning in kernel mode on the master kernel lock we
have for SMP. Userland is of course multi threaded on SMP, but the
kernel is simgle threaded so all those faults being done by your test
program must be processed serially.

Yow! 11.26 MB/s remote host TCP bandwidth & ////
199 usec remote TCP latency over 100Mb/s ////
ethernet. Beat that! ////
-----------------------------------------////__________ o
David S. Miller, /_____________/ / // /_/ ><