Re: dentries rule. [was: Re: Pros and cons of newer kernels]

Steven N. Hirsch (shirsch@ibm.net)
Tue, 18 Nov 1997 20:45:12 -0500 (EST)


On Tue, 18 Nov 1997, Ingo Molnar wrote:

>
> On Tue, 18 Nov 1997, Steven N. Hirsch wrote:
>
> > Have _you_ made such a comparison? I have found that /usr/man/man3 is a
> > great target for the test. Mine is > 70k and has ~2200 files in it.
>
> yep:
>
> 'time mc+F10+Enter' in /usr/man/man3 (885 files) on a 32M 100 MHZ P5
> system, cached case, non-SMP kernel:
>
> 2.0.31: 0.32 user 0.99 system 0:01.43 elapsed 91%CPU
>
> 2.1.64: 0.28 user 0.11 system 0:00.48 elapsed 81%CPU
>
> ie. under 2.1, Midnight Commander starts up 3 times faster, time spent in
> the kernel is _ten times lower_, ie: dentries rule. The same is true for
> 'pine' and basically all other applications i use and where interactive
> performance matters.

Ah, but when I do that:

2.0.31: 0.72user 5.11system 0:06.19elapsed 94%CPU (486/133)

2.1.64: 0.32user 0.57system 0:17.04elapsed 5%CPU (6x86 P150+)

Just a bit of difference in the CPU usage, no? Any suggestions as to why
the 6x86 shows so little CPU use? Top shows the system as being 97%
idle! Could this be some odd misbehavior with the Cyrix idle patch?

Very strange.

Steve