Re: vm lock contention reduction

Rik van Riel wrote:

quoted

But it is, mmap() and anonymous memory don't trigger writeback.

On Fri, Jul 05, 2002 at 05:48:48PM -0700, Andrew Morton wrote:

That's different.  Bill hit a problem just running tiobench.
We can run balance_dirty_pages() when a COW copyout is performed,
which will approximately improve things.
But the whole idea of the dirty memory thresholds just seems bust,
really.  Because how do you pick the thresholds?  40%.  Bah.

I don't know what the answer should be, but I can certainly demonstrate
this in a rather uninteresting situation (4GB, 4cpu's, 1 disk, 16 tasks).

But I can concur with that evaluation. In my esteem fixed fractions of
memory don't have a very direct relationship to what's going on.


Cheers,
Bill
--
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/