From: "Tony Ingenoso" 

That only softens the blow - L2 hits are still far more costly than L1
hits. It helps some with the bloat and in multi-tasking environmants, but
its still dramatically slower than if the code/data were mostly living in
the L1.

Its really just admitting that code sucks and there's nothing we can do
about it 

"Geo"  wrote in message
news:427b3eca$1{at}w3.nls.net...
> I've noticed the more recent specs are showing 2mb L2 cache levels. So it
> looks like they are going for a hardware solution to this issue.
>
> Geo.
>
> "Tony Ingenoso"  wrote in message
> news:427aefc5{at}w3.nls.net...
> > That works fine until some programmer tells their compiler to
"optimize"
> for
> > speed and the code porks up due to loop unrolling, function inlining and
> > such.  Now you can be in a situation where the stuff is so large its not
> > only blowing the L1, its blowing the L2 as well and using extra pages.
> >
> > Having to fault in or swap out an extra page can negate all the nominal
> > speed gains you may have gotten by producing "book fast" code.
>
>
>

--- BBBS/NT v4.01 Flag-5