[Arm-netbook] RedSleeve - New ARM Linux Distro

[Gordan Bobic]

cnxsoft wrote:
> On 13/04/2012 04:53, Gordan Bobic wrote:
>> On 12/04/2012 22:27, Henrik Nordström wrote:
>>> mån 2012-03-26 klockan 09:38 +0100 skrev Gordan Bobic:
>>>
>>>> Have you actually looked at what the realistic performance advantage is
>>>> on something like a Tegra2 between running armv5tel and armv7hl for a
>>>> typical desktop workload?
>>> I have not tried them in desktop setup, but in batch compile performance
>>> difference there is a very noticeable difference between Tegra2 running
>>> armv5tel or Tegra2 running armv7hl for any larger compile job even when
>>> running the same kernel.
>> Can you put a number on that over a few different tasks, e.g. a 100
>> loops of bzip2 of a 100MB file generated from /dev/random going to
>> /dev/null and 100 iterations of compiling something small in tmpfs?
>>
>> "Very noticeable" isn't very scientific. :)
>>
>> Gordan
>>
> Phoronix compared Ubuntu 11.10 (soft-float) against Ubuntu 12.04 
> (hard-float) on the Pandaboard.
> http://www.phoronix.com/scan.php?page=article&item=ubuntu_1204_armhf&num=3
> 
> This is also not very scientific, as other optimizations may have 
> improved performances, but still worth a look.

It is also not even clear that 12.04 is hard-float - both list kernels 
as armv7l, rather than armv7hl. The test where I'd have expected to see 
no difference (gzip compression isn't exactly very FP heavy) has one of 
the biggest differences. So I'm not convinced how relevant the results 
are to a relative performance assessment of SF vs. HF - at a glance it 
seems more plausible the differences come from optimizations in the 
packages themselves and possibly a better compiler toolchain.

Gordan