Large pages, large blocks, and large problems
Posted Sep 19, 2007 16:22 UTC (Wed) by james
Parent article: Large pages, large blocks, and large problems
Some of Linus' thoughts (presumably) can be found at
Real World Technologies (and associated thread):
[The performance cost of] Page table handling stays pretty constant - you basically
get TLB misses proportionately to your data size, which
means that the more TLB misses you get, the more data cache
misses you get!
So realistically, TLB costs are never going to grow in
any unbounded kind of manner - they are always limited by
(and generally much smaller than) the D$ costs!
There are loads that are more TLB-intensive than others
(and loads that are more D$ intensive, of course!), but
in the end, TLB's aren't the problem.
Unless the CPU micro-architecture is unbalanced, of course.
There have certainly been uarchs that increased the cache
size a lot without increasing the TLB size. Now, of
course they'll be TLB-limited! But that's not really a
fundamental issue, it's just an unbalanced design.
You want good cache behavior if you have 256GB of memory,
or your performance will suck. It's that easy. And if you
have good locality in the D$, then the TLB's will work
to post comments)