Re: [gentoo-user] OT: but cool - NASDAQ is gentoo powered

[James Broadhead]

On 18 August 2011 09:23, Norman Rieß <norman@×××××××××.org> wrote:
> Am 08/18/11 09:11, schrieb Matthew Finkel:
>> Just out of curiosity, how long does it take to compile gcc?
>>
>> - Matt
>
> Atom:
>
> genlop -t sys-devel/gcc-4.4.5
>  * sys-devel/gcc
>
>     Sat Feb 26 13:06:08 2011 >>> sys-devel/gcc-4.4.5
>       merge time: 1 hour, 12 minutes and 27 seconds.
>
>     Wed Mar 23 23:01:12 2011 >>> sys-devel/gcc-4.4.5
>       merge time: 1 hour, 10 minutes and 22 seconds.

I have an Atom 330 machine which is getting significantly worse
build-times than you. What make.conf options are you using? (Or are
you using something else to improve build times?)

     Wed Mar 16 04:49:09 2011 >>> sys-devel/gcc-4.4.5
       merge time: 2 hours, 56 minutes and 20 seconds.

     Thu May  5 22:07:36 2011 >>> sys-devel/gcc-4.3.4
       merge time: 2 hours, 14 minutes and 15 seconds.

     Fri May  6 00:35:53 2011 >>> sys-devel/gcc-4.4.5
       merge time: 2 hours, 28 minutes and 17 seconds.

Admittedly, my machine runs xbmc, which is a resource hog, and has a
fair bit of disk activity.
My CFLAGS are:
CFLAGS="-O2 -march=core2 -mtune=generic -fomit-frame-pointer -pipe
-mssse3 -mfpmath=sse"
which date to before -march=atom, and having read a performance
article suggesting these. I note that the only practical difference
between the resultant gcc options is that setting -mtune to core2 adds
"#define __tune_core2__ 1". I wonder what the practical difference is.
echo | gcc -dM -E - -O2 -march=core2 -mtune=generic
-fomit-frame-pointer -pipe -mssse3 -mfpmath=sse

I suppose, having looked into it this far, I'll merge gcc-4.5 to see
what effect -mtune=atom has.

(I'm not particularly interested in build times, but whether they're a
sign of poor overall performance ... )

JB