On Sun, 8 Nov 1998, MOLNAR Ingo wrote:
> can anyone write some code that clearly demonstrates that we lose
> performance on SMP?
I once patched the kernel a bit to measure when exactly scheduling
happens and how processes switch processors. It especially showed
unfairnes with niced processes (I didn't measure performance loss).
I made two traces available:
ftp://nice.ethz.ch/users/felix/nice_1.xg.gz
ftp://nice.ethz.ch/users/felix/nice_3.xg.gz
They show one and three processes on a two processor machine, all
processes are niced to 20. The two files are xgraph files (just ungzip
them and load them with xgraph). The second trace shows that the
last started job only gets the CPU when the other two finished. The
kernel I used was 2.0.3x (ask me if you need the details).
- Felix
--
Felix Rauch | Email: [EMAIL PROTECTED]
Institute for Computer Systems | Homepage: http://www.cs.inf.ethz.ch/~rauch/
ETH Zentrum / RZ H15 | Phone: ++41 1 632 7489
CH - 8092 Zuerich / Switzerland | Fax: ++41 1 632 1307
