Re: 2.6.5-mm5

[Nick Piggin <nickpiggin@yahoo.com.au>]

Kurt Fitzner wrote:
> Andrew Morton wrote:
> 
>>
>> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mm5/ 
>>
>>
>> - More CPU scheduler work.  Hopefully this kernel will now address the
>>   regressions that a few people have noted on certain workloads.  We 
>> appear to
>>   be getting close.
> 
> 
> Regressions... from the context, I'm assuming you're not talking
> regression errors here?  I'm assuming these are performance issues?  I
> see a 3.5% drop in compiling speed between 2.4.24 and 2.6.5 on a dual
> Athlon workstation.  I'll test this kernel happily if the scheduler
> tweaks are intended to address this.
> 

Hi Kurt,
The context was actually sched-domains regressions vs numasched,
which might possibly arise in any SMP (even simple dual) system.
So in this case we are interested in -mm regressions compared to
the official 2.6 tree.

2.6 regressions versus 2.4 are still interesting, but a 3.5% drop
in kernel compiling is probably due to HZ=1000 and rmap, although
I think you can expect improvements in rmap overhead soon... try
2.6.5-aa5 if you are interested.