On 12/08/2012 08:38 PM, M A Young wrote:
On Fri, 7 Dec 2012, Brendan Conoboy wrote:
If you could run a few benchmarks between the v6 and v5 spins we'd all
be *very* interested to know the difference in performance. Even
simple stuff like time to bzip2 a cached file would be informative.
Here are some very basic benchmark stats
time bzip2 -9 <initramfs-3.6.9-2.fc17.x86_64.img >/dev/null
[v6]
real 1m7.734s
user 1m7.540s
sys 0m0.160s
[v5]
real 1m10.920s
user 1m10.650s
sys 0m0.210s
time md5sum initramfs-3.6.9-2.fc17.x86_64.img >/dev/null
[v6]
real 0m0.431s
user 0m0.290s
sys 0m0.130s
[v5]
real 0m0.433s
user 0m0.270s
sys 0m0.150s
time sha256sum initramfs-3.6.9-2.fc17.x86_64.img >/dev/null
[v6]
real 0m1.393s
user 0m1.240s
sys 0m0.150s
[v5]
real 0m1.436s
user 0m1.350s
sys 0m0.080s
An earth shattering difference, then.
This is the second time now that it has been demonstrated than having a
armv6hl distro is a waste of time. Somebody rebuilt Debian for the Pi
only in that case it was a difference between armel (armv4) and armv6hl,
and the difference was similarly negligible.
http://www.memetic.org/raspbian-benchmarking-armel-vs-armhf/
The only place where the difference is meaningful in their benchmarks
was in AV transcoding which isn't a particularly obvious use-case for a
Pi. The difference between armv5tel (what Fedora soft-float targets) and
armv6hl is smaller.
Gordan
_______________________________________________
arm mailing list
arm@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/arm
