login
Header Space

 
 

Re: [Patch](memory hotplug) Make kmem_cache_node for SLUB on memory online to avoid panic(take 3)

Score:
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: Andrew Morton <akpm@...>
Cc: Yasunori Goto <y-goto@...>, Linux Kernel ML <linux-kernel@...>, linux-mm <linux-mm@...>
Date: Thursday, October 18, 2007 - 5:13 am

On Thu, 18 Oct 2007, Andrew Morton wrote:


Well the problem right now is the regression in slab_free() on SMP. 
AFAICT UP and NUMA is fine and also most loads under SMP. Concurrent 
allocation / frees on multiple processors are several times faster (I see 
up to 10 fold improvements on an 8p).

However, long sequences of free operations from a single processor under 
SMP require too many atomic operations compared with SLAB. If I only do 
frees on a single processor on SMP then I can produce a 30% regression for 
slabs between 128 and 1024 byte in size. I have a patchset in the works 
that reduces the atomic operations for those.

SLAB currently has an advantage since it uses coarser grained locking. 
SLAB can take a global lock and then perform queue operations on 
multiple objects. SLUB has fine grained locking which increases 
concurrency but also the overhead of atomic operations.

The regression does not surface under UP since we do not need to do 
locking. And it does not surface under NUMA since the alien cache stuff in 
SLAB is reducing slab_free performance compared to SMP.

-
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[PATCH] Fix warning in mm/slub.c, Olof Johansson, (Tue Oct 23, 12:21 am)
Re: [PATCH] Fix warning in mm/slub.c, Christoph Lameter, (Tue Oct 23, 12:21 pm)
Re: [PATCH] Fix warning in mm/slub.c, Yasunori Goto, (Tue Oct 23, 1:35 am)
Re: [PATCH] Fix warning in mm/slub.c, Pekka Enberg, (Tue Oct 23, 3:52 am)
Re: [Patch](memory hotplug) Make kmem_cache_node for SLUB on..., Christoph Lameter, (Thu Oct 18, 2:25 am)
Re: [Patch](memory hotplug) Make kmem_cache_node for SLUB on..., Christoph Lameter, (Thu Oct 18, 5:13 am)
speck-geostationary