[clue-tech] 2.6 kernel compile options make Intel CPU slow

[Jim Ockers]

Hi,

Maxwell Spangler wrote:
> On Mon, 2010-07-19 at 15:32 -0600, Jim Ockers wrote:
>   
>> Hi CLUEbies,
>>
>> I have a curious problem.  We compiled a custom 2.6 Linux kernel for
>> an embedded platform.  The target hardware is Intel P4 or AMD.  The
>> hardware used to run a 2.4 Linux kernel.  Here is the fast/slowness
>> matrix:
>>
>>        Intel      AMD
>> 2.4    fast       fast
>> 2.6    slow       fast
>>     
>
> This small bit of data suggests that the 2.6 kernel can run as fast as
> the 2.4 but that something about the 2.6 + Intel P4 hardware combination
> is slow.
>   
Yes, that is exactly what we have determined.
> Can you run the same test on a variety of Intel and AMD chips to see if
> a pattern develops?  There might be other AMD chips that suffer as well
> as other Intel chips that are successful.  Comparing the differences in
> chip features might point to why your results here are problematic.
>   
I think we've established the pattern because it seems to affect all 
Intel P4s running that particular kernel. When we use a stock CentOS 
kernel (which doesn't have our custom IDE and network card driver 
hacks^H^H^H^H^Hmodifications and other custom settings, then that 
hardware is "fast." When we use our custom kernel, it's slow only on 
Intel P4.
> Also, were there any clues in the results of the time command's output?
> Any big differences on the level of what time was being used: user vs
> sys?
>
>   
Regardless of how long the dd test takes to return, or if the system is 
"fast" or "slow," it is almost all sys time that is used. The "user" 
time is negligible.

Jim

-- 
Jim Ockers, P.Eng. (ockers at ockers.net)
Contact info: http://www.ockers.ca/pason.html


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cluedenver.org/pipermail/clue-tech/attachments/20100719/c0553790/attachment.html