logo       

variability in Oprofile sampling: msg#00038

Subject: variability in Oprofile sampling
Hi all,
I'm getting some strange samples when I try the same program on different systems but the same Linux OS.

System #1
AMD Opt 2.5GHz  1 processor

System #2
Intel P4 3.6 Ghz    2 cores

System #3
AMD Opt 2.0 Ghz  2 cores

Repeated runs on the same machine produce the same results.

So the top three samples for Linpack result is vastly different profiles. 
                  1st                          2nd                              3rd
System #1  36.22% libat.so        35.64% Linpack       10.20% ps.so

System #2  61.39% no-vmlinux   13.86 % Linpack       15.16% libat.so

System #3 34.19% Linpack        27.96% libat.so         19.82 no-vmlinux

The runtimes for these programs is nearly identical! However the samples are very different.  vm-linux is reported for #2 and #3, but #1 does not have any no-vmlinux samples!

As can be seen, the program has a fair amount of dynamic link library usage.  Could this be a factor for the very different profiles?  Also can hardware make such a huge difference, somehow affecting how the sample is done?  As can be seen System #1 and #3 vs #2 are very different. 

Thanks,
-Wei
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
oprofile-list mailing list
oprofile-list@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/oprofile-list
<Prev in Thread] Current Thread [Next in Thread>