At Thu, 25 May 2006 18:21:39 +0000 Alex <[EMAIL PROTECTED]> wrote:
> Allan Gottlieb wrote:
>> At Thu, 25 May 2006 10:40:26 +0000 Alex <[EMAIL PROTECTED]> wrote:
>>
>>>Allan Gottlieb wrote:
>>>
>>>>You have to do experiments. It depends heavily on your application
>>>>mix.
>>>
>>>Yes, that would be the best, but I'm wondering how, because e.g. "time
>>>bzip2 -9 foobar" wouldn't be helpfull. So now I've switched to "-Os"
>>>and soon I can test, if it's a real difference.
>> Please report back your findings, including the application mix you
>> tested. Although "scientific" timed benchmarks are important, I would
>> also be interested in how the system feels. For the latter ("feels"),
>> you should qualitatively describe the use of the system (web server,
>> desktop, laptop, etc) and what you commonly run (program devel, games,
>> scientific/engineering apps, etc).
>> thanks,
>> allan
>
> Hi,
>
> I've a desktop system and I commonly use applications like firefox,
> thunderbird and so on, kde, gaim and a terminal is nearly always
> there. Sometimes I'm running vim or kate.
>
> If you're interested in some tests, not relevant for desktop systems,
> there are some I made:
>
> Time wasted to compress a 416 mb tar:
> bzip2 gzip
> -O3 2m40.882s 1m20.445s
> -Os 2m39.314s 1m21.157s
>
> decompress:
> bzip2 gzip
> -O3 0m52.575s 0m4.972s
> -Os 0m53.387s 0m4.828s
>
> Convert 203 Mbs MP3s to WAV using LAME:
> -O3 14m4.461s
> -Os 16m50.599s
>
> from wav to mp3:
> -O3 1m1.708s
> -Os 1m12.841s
>
> Now I'm emerging -e world with -Os. When it is finished, I'll mail you
> the results.
The conversion programs you ran might not stress the memory system. I
suspect that they only keep a fixed size portion of the input and
output files in memory when you run them with ever larger inputs.
allan
--
[email protected] mailing list
