Optimized kernel builds: the straight dope

[Hervé Fache]

I am a bit surprised nobody commented on these results: they seem to
make a case for -k7 kernels in some configurations...

Did I completely waste my time?

Hervé.

On 8/27/06, Hervé Fache <Herve at lucidia.net> wrote:
> Hello all,
>
> Took me too long to find the time! But never mind, here are the results:
>
> no_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    335     97.3    0.0     0.0     1.00
> 2.6.15-26-k7       3    330     98.2    0.0     0.0     1.00
> cacherun:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    326     99.1    0.0     0.0     0.97
> 2.6.15-26-k7       3    326     99.1    0.0     0.0     0.99
> process_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    472     68.9    504.0   30.5    1.41
> 2.6.15-26-k7       3    467     69.2    444.3   29.6    1.42
> ctar_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    418     79.2    6.3     4.8     1.25
> 2.6.15-26-k7       3    394     84.0    6.0     4.8     1.19
> xtar_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    622     54.0    14.3    11.6    1.86
> 2.6.15-26-k7       3    535     62.2    12.3    12.0    1.62
> io_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    872     38.2    13126.4 18.8    2.60
> 2.6.15-26-k7       3    820     40.6    12280.1 19.6    2.48
> read_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    1196    28.0    86.9    6.7     3.57
> 2.6.15-26-k7       3    510     64.9    26.7    4.7     1.55
> list_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    429     76.9    1.3     12.6    1.28
> 2.6.15-26-k7       3    413     79.9    1.0     12.8    1.25
> mem_load:
> Kernel        [runs]    Time    CPU%    Loads   LCPU%   Ratio
> 2.6.15-26-386      3    354     91.8    42.7    0.0     1.06
> 2.6.15-26-k7       3    340     95.3    42.3    0.0     1.03
>
> As you can see, all similar number apart from the read_load test. We
> still seem to benefit from the optimized kernel under load though.
>
> HTH
> Hervé.
>
> On 8/15/06, Hervé Fache <Herve at lucidia.net> wrote:
> > Matt,
> >
> > I would like to run this test on my AthlonXP 2800+ box, which is a
> > very fast machine. This might bring some interesting results from the
> > -k7 side?
> >
> > Hervé.
> >
> > On 8/14/06, Matt Zimmerman <mdz at ubuntu.com> wrote:
> > > Enter Ben Collins, our kernel team lead, who has put together a performance
> > > test to answer that question, covering both the i386 and amd64
> > > architectures.
> >
>


-- 
In a world without walls and fences, who needs Windows and Gates?