On Mon, Jun 27, 2011 at 03:19:43PM +0200, Arnd Bergmann wrote:

quoted

quoted

I suppose for the majority of the cases, the overhead of the indirect
function call is near-zero, compared to the overhead of the cache
management operation, so it would only make a difference for coherent
systems without an IOMMU. Do we care about micro-optimizing those?

FWIW, when I was hacking on ARM access point routing performance some
time ago, turning the L1/L2 cache maintenance operations into inline
functions (inlined into the ethernet driver) gave me a significant and
measurable performance boost.

Such things can remain product-specific hacks, though.

quoted

Even in coherent case, the overhead caused by additional function call
should have really negligible impact on drivers performance.

What about object code size? I guess since ixp23xx is the only platform
that announces itself as coherent, we probably don't need to worry about
it too much either. Lennert?

I don't think so.  ixp23xx isn't a very popular platform anymore either,
having been discontinued some time ago.


thanks,
Lennert