On Wed, 9 Feb 2000, Linus Torvalds wrote:
> On Wed, 9 Feb 2000, Andrea Arcangeli wrote:
> > It basically avoids the caching part of the icache and dcache to grow more
> > than 2% of the memory in the machine.
>
> Heuristics like these are BAD. The heuristicsfor trying to limie buffer
> cache and page cache were one of the reasons that we tended to suck at
> specweb, because it silently just didn't use available memory for caching.
>
> Don't limit what the memory is used for - let the dynamics of the
> allocators do the best they can. If there are problems with the
> _freeing_ of the memory, that is what should be fixed, not this.
the problem andrea was trying to address is that there is no impedance to
cache growth on large memory hardware. letting the caches grow until they
compete with each other for physical memory works well on small memory
hardware. but on large memory hardware, large caches pollute the CPU
caches during searches, actually slowing the system down.
a fixed limit is probably not the ultimate answer. but somehow the caches
need other forces on them to keep them sized reasonably.
- Chuck Lever
-- corporate: <chuckl@netscape.com> personal: <chucklever@netscape.net> or <cel@monkey.org>The Linux Scalability project: http://www.citi.umich.edu/projects/linux-scalability/
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 15 2000 - 21:00:17 EST