Re: [patch 1/3] slob: rework freelist handling

[Nick Piggin <npiggin@suse.de>]

On Wed, May 23, 2007 at 08:49:30PM -0700, Christoph Lameter wrote:
> On Thu, 24 May 2007, Nick Piggin wrote:
> 
> > The reason SLOB is so space efficient really comes from Matt's no
> > compromises design. The thrust of my patches were after seeing how slow
> > it was on my 4GB system while testing the RCU implementation. They
> > were primarily intended to speed up the thing, but retain all the same
> > basic allocation algorithms -- a quirk of my implementation allowed
> > smaller freelist indexes which was a bonus, but as Matt said, slob was
> > still more efficient before the change.
> 
> Well as far as I understand Matt it seems that you still need 2 bytes per 
> alloc. That is still more than 0 that SLUB needs.

That's true, but I think the more relevant number is that SLUB needs
400K more memory to boot into /bin/bash.

  
> > What SLUB idea did you think I copied anyway?
> 
> The use of the page struct.

The general use of struct page to store page metadata other than pagecache?
I guess not, because that predates SLUB and even SLAB. You've got a
freelist pointer in there, so maybe that's what you mean. I was hoping
for something that I actually can "steal" to make SLOB even better ;)

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>