Re: [RFC] [PATCH] virtual memmap on sparsemem v3 [0/4] introduction

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Sun, 10 Dec 2006 14:47:30 -0500
"Bob Picco" <bob.picco@hp.com> wrote:
> > Intro:
> > When using SPARSEMEM, pfn_to_page()/page_to_pfn() accesses global big table
> > of mem_section. if SPARSEMEM_EXTREME, this is 2-level table lookup.
> Did you gather any performance numbers comparing
> VIRTUAL_MEM_MAP+SPARSEMEM to SPARSEMEM+EXTREME? I did some quick but
> inconclusive (small machine) ones when you first posted. There was
> perhaps a slight degradation in VIRTUAL_MEM_MAP+SPARSEMEM.
> 
No, I didn't. I'll do when I have a chance to do it. I hope that this won't
be merged until someone shows the benefit by data.
(I think this verion is better than the first one but..)

IIRC, DISCONTIGMEM + VIRTUAL_MEM_MAP shows a bit better performance than
SPARSEMEM_EXTREME. What I expect is to archive VIRTUAL_MEM_MAP performance
with SPARSEMEM.

Now, we still have chance to optimization.
- optimize pfn_valid. (I'll post this. mem_section[] will be never accessed
  in runtime by this.)
- use large-sized-page (no concrete idea, maybe to modify map/unmap func
  will work enough.)

Thanks,
-Kame

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