Re: [RFT/PATCH] slab: consolidate allocation paths

[Pekka Enberg <penberg@cs.helsinki.fi>]

Hi,

On Sat, 2006-02-04 at 18:00 -0800, Paul Jackson wrote:
> Two issues I can see:
> 
>   1) This patch increased the text size of mm/slab.o by 776
>      bytes (ia64 sn2_defconfig gcc 3.3.3), which should be
>      justified.  My naive expectation would have been that
>      such a source code consolidation patch would be text
>      size neutral, or close to it.

Ah, sorry about that, I forgot to verify the NUMA case. The problem is
that to kmalloc_node() is calling cache_alloc() now which is forced
inline. I am wondering, would it be ok to make __cache_alloc()
non-inline for NUMA? The relevant numbers are:

   text    data     bss     dec     hex filename
  15882    2512      24   18418    47f2 mm/slab.o (original)
  16029    2512      24   18565    4885 mm/slab.o (inline)
  15798    2512      24   18334    479e mm/slab.o (non-inline)

>   2) You might want to hold off this patch for a few days,
>      until the dust settles from my memory spread patch.

Sure.

			Pekka