On http://developer.osdl.org/craiger/hackbench one finds hackbench.c
and the result that 2.6 is much better than 2.4 where scheduling is
concerned.
I was shown results that go in the other direction, so just tried
on a machine here.
50 processes
2.4.26: 24 sec
2.6.0t11: 72 sec
2.6.8.1: 120 sec
20 processes
2.4.26: 8.6 sec
2.6.0t11: 29 sec
2.6.8.1: 30 sec
This was on a 256 MB 400 MHz Pentium II.
Andries
El Sat, 18 Sep 2004 20:07:18 +0200 (MEST) <[email protected]> escribi?:
> I was shown results that go in the other direction, so just tried
> on a machine here.
In a 512 MB dual p3 machine I got this: (hackbench 100)
2.4.26: Time: 38.123
2.6.9-rc2-mm1: Time: 19.505
BTW, in the 2.6 case the kernel didn't have a very nice behaviour: A couple of
seconds after starting hackbench(after doing the benchmarks), Xorg stopped
running until hackbench finished. X was using nice 0.