lkml.org 
[lkml]   [2011]   [Jul]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] slub: reduce overhead of slub_debug
On 07/07/2011 11:42 AM, Christoph Lameter wrote:
> On Thu, 7 Jul 2011, Ben Greear wrote:
>
>> The more painful you make it, the less likely folks are to use it
>> in environments that actually reproduce bugs, so I think it's quite
>> short-sighted to reject such performance improvements out of hand.
>>
>> And what if some production machine has funny crashes in a specific
>> work-load....wouldn't it be nice if it could enable debugging and
>> still perform well enough to do it's job?
>
> Sure if there would be significant improvements that accomplish what
> you claim above then that would be certainly worthwhile. Come up with
> patches like that please.

The patch appears to make some work loads twice as fast ('make clean'),
and it had a reasonable speedup to the 'make -j12'. What do you
consider 'significant'?

I'm willing to do some other network-related benchmarks with his patch if
that would give it better chance of being accepted. (I end up running
with SLUB debug quite a bit on big, heavy, workloads...so any speedup
in that would be a big help for us...)

Thanks,
Ben

--
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc http://www.candelatech.com



\
 
 \ /
  Last update: 2011-07-07 20:57    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site