Re: Note describing poor dcache utilization under high memory pressure

[Daniel Phillips]

On January 29, 2002 12:12 am, Rick Stevens wrote:
> Daniel Phillips wrote:
> 
> > Since the page was copied to the child, the child's page table must be 
> > altered, and since it is shared, it must first be instantiated by the child.  
> > So after all the dust settles, the parent and child have their own copies of 
> > a page table page, which differ only at a single location: the child's page 
> > table points at its freshly made CoW copy, and the parent's page table points 
> > at the original page.
> 
> > 
> > The beauty of this is, the page table could just as easily have been shared 
> > by a sibling of the child, not the parent at all, in the case that the parent 
> > had already instantiated its own copy of the page table page because of an 
> > earlier CoW.
> 
> 
> Ok.  Still seems like a bit more copying than necessary.

I think I can show that it's exactly as much copying as necessary, and no more.
Oh, the page directory could also be shared and, on a 3 level page table, so
could the mid level tables, but that's not a really big win because of the 1K/1
fanout.  I.e, we already took care of 99.9% of the problem just by sharing the
bottom-level page tables.

> I'd have to look at it a bit more and do some noodling.
>  
> > Confused yet?  Welcome to the club ;-)
> 
> Does my head exploding qualify for "confused"?  If so, then I'm not
> yet "confused".  I'm "concerned", since my ears are bleeding (a
> precursor to an explosion) ;-p

:-)

-- 
Daniel