Re: CPU specific/optimized Debian builds ?
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Wednesday 29 May 2002 12:08 pm, Oliver Kurth wrote:
> If bzip2 should be used as a benchmark, write a small program that
> generates random data and pipe this to bzip2:
>
> genran | bzip2 -c > /dev/null
>
> This should not generate any disk IO. Of course, the data generating
> program should not be CPU intensive itself.
Tests use a file generated by dd if=/dev/urandom of=./test.file bs=1M count=10
Compressed with bzip2 -kc ./test.file > /dev/null
File is in cache, so no disk io.
gcc-2.95.4 (no optimisation)
19.1 seconds
gcc-2.95.4 (-O2 -fomit-frame-pointer)
15.8 seconds
gcc-2.95.4 (-O2 -fomit-frame-pointer -march=i686)
16.1 seconds
gcc-3.0 (no optimisation)
24.75 seconds
gcc-3.0 (-O2 -fomit-frame-pointer)
16.0 seconds
gcc-3.0 (-O2 -fomit-frame-pointer -march=i686)
15.9 seconds
gcc-3.1 (no optimisation)
22.1 seconds
gcc-3.1 (-O2 -fomit-frame-pointer)
15.6 seconds
gcc-3.1 (-O2 -fomit-frame-pointer -march=pentium4)
15.8 seconds
Summary:
gcc-3.1 is the fastest compiler (even faster than icc). Turning on
cpu-specific optimisation decreses performance in this case; I'm not sure if
that is noise (performance goes up for gcc-3.0), but the results were pretty
consistent.
So there it is... in this case, at least, cpu-specific optimisations don't
appear to be worth much. Of course, that might not be true for different
workloads.
*matt*
- --
Now I know someone out there is going to claim, "Well then, UNIX is intuitive,
because you only need to learn 5000 commands, and then everything else follows
from that! Har har har!"
-- Andy Bates on "intuitive interfaces", slightly defending Macs
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iD8DBQE89SoQoVQMDIAmueURAgTpAJ0bP7imcJuy/xFlSoFpv52WRQUHawCfZvSM
fFvm0OA2HVfa0mA2z802gGw=
=EiGE
-----END PGP SIGNATURE-----
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: