Re: [PATCH] Shrinks sizeof(files_struct) and better layout

[Andi Kleen]

On Wednesday 04 January 2006 12:41, Eric Dumazet wrote:

> > The overhead of the kmem_cache_t by itself is negligible.
> 
> This seems a common misconception among kernel devs (even the best ones Andi :) )

It used to be true at least at some point :/

> 
> On SMP (and/or NUMA) machines : overhead of kmem_cache_t is *big*
> 
> See enable_cpucache in mm/slab.c for 'limit' determination :
> 
>          if (cachep->objsize > 131072)
>                  limit = 1;
>          else if (cachep->objsize > PAGE_SIZE)
>                  limit = 8;
>          else if (cachep->objsize > 1024)
>                  limit = 24;
>          else if (cachep->objsize > 256)
>                  limit = 54;
>          else
>                  limit = 120;
> 
> On a 64 bits machines, 120*sizeof(void*) = 120*8 = 960
> 
> So for small objects (<= 256 bytes), you end with a sizeof(array_cache) = 1024 
> bytes per cpu

Hmm - in theory it could be tuned down for SMT siblings which really don't
care about sharing because they have shared caches. But I don't know
how many complications that would add to the slab code.

> 
> If 16 CPUS : 16*1024 = 16 Kbytes + all other kmem_cache structures : (If you 
> have a lot of Memory Nodes, then it can be *very* big too).
> 
> If you know that no more than 100 objects are used in 99% of setups, then a 
> dedicated cache is overkill, even locking 100 pages because of extreme 
> fragmentation is better.

A system with 16 memory nodes should have more than 100 processes, but ok.

> 
> Maybe we can introduce an ultra basic memory allocator for such objects 
> (without CPU caches, node caches), so that the memory overhead is small. 
> Hitting a spinlock at thread creation/deletion time is not that time critical.

Might be a good idea yes. There used to a "simp" allocator long ago for this,
but it was removed because it had other issues. But this was before even slab
got the per cpu/node support.

-Andi
 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/