[En-Nut-Discussion] RFC: New driver interface for TWI, I2C and others

[Ulrich Prinz]

Hi Ole,

Am 03.05.2011 16:15, schrieb Ole Reinhardt:
> Hi all,
> 
> No I'm not yet fed up :)

Great to hear, but I never thought that, as I know that business can
sometimes be hard and leaves not time.
> 
> I'm still the same opinion like Ulrich, but unfortunately this is quite
> a big change. But it should be possible to handle this without big
> impacts on the "normal" application code (at least for the UART
> drivers).

Ah, no... It isn't... I implemented this way of USART setup on the
devnut_m3n branch for Cortex. The driver API has not changed. It still
looks like and is called like the good old AVR code. There are some
ioctl() calls added that optionally enable some features but they only
would return an error if called with AVR. Thats all.
> 
> The bus architecture you implemented for SPI on the ARM architecture is
> realy a big step forward and I think we should implement TWI/I2C the
> same way.

What implementation do you mean, Haralds, mine or the combination of
them both?

But yes it is a change as it should be implemented for all ARM
architectures. They are fast enough and have the needed bytes of RAM
that an AVR cannot spare. But TWI implementation shows, that a bridge
between AVR and ARM can be kept that allows to use old API functions and
nutconf defaults in demo code.
> 
> Ulrich: Perhaps we could start into a discussion by a private
> conversation? I think we should work on these changes in a development
> branch. Might be good to have these changes for the 5.0 release?
> 
These private discussions are sometimes very good as things can be taken
fast and without interrupts and long delays where you have forgotten the
good idea from yesterdays email :)

Email or phone call any time you like this week?

Then we put some things straight and come back to here, fighting the
same direction :)

Ulrich