On Wed, Sep 18, 2013 at 10:40:22AM -0500, David Woodhouse wrote:
On Tue, 2013-09-17 at 20:13 +0100, Mark Brown wrote:
quoted
quoted
[ potential "LUT" optimization, implementation details ]
quoted
AFAICT the big issue with the LUTs in this controller is that they are
expensive to reprogram so things are going to work better if there is a
small set of known operations that will happen repeatedly. Otherwise
walking the transfer list at runtime isn't too hard.
I think you can drop the word "known" there. With decent management of
the LUT as a kind of LRU cache of recent operations, surely you'll
quickly get into a situation where the operations you're actually
*doing* on the flash are all already in the LUT and you're never
actually *having* to reprogram it?
Yeah, assuming it's not too annoying to identify the repeated patterns
and doesn't blow up horribly just caching should work fine.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20130918/41f92132/attachment.sig>