On Monday 14 April 2008, Jean Delvare wrote:
> > > To support the X24645, it would be necessary to raise AT24_MAX_CLIENTS
> > > to 32 (what a beast!). Then again, most eeproms will just need one
> > > client, so this would cause quite some overhead in most use-cases.
> > > Maybe it pays off to hande this dynamically?
> > 
> > As eeproms are normally slow things anyway, would it be a big
> > performance impact?
> 
> I agree that going dynamic makes sense. Not that it has anything to do
> with the speed of EEPROMs though - the memory allocation will be done
> at device initialization time, and after that dynamic or not should
> perform just the same.

Right.  X24645 looks a bit more bizarre than any EEPROMs
I had come across; it'll need some changes in chip config
data.  You'd think they'd just use two byte addressing!
I guess doing it that way lets them use SMBus-only hosts.


Wolfram's at24_ee_address() cleanup looked fine.

The header looks OK, but I'd add a comment *encouraging*
folk to double check those params for their chips, in
particular to check the page size.  It's OK if their chip
has a bigger page size than those "defaults" although they
could get faster write performance by listing the right
value ... but it's NOT OK if their chip has a smaller write
size than what's listed there, they'll lose data.


I'll be glad to see this go upstream.  :)

- Dave

_______________________________________________
i2c mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/i2c

Reply via email to