Re: [libopencm3-devel] Deduplication

[Ken S. <ksa...@in...>]

Hi Chrysn

On 18/10/12 02:36, chrysn wrote:
> for peripherials, there are three different topics as far as i am
> concerned:
>
> * regular peripherials
>
>    peripherial code will not be exchangable in general. for the efm32
>    platform, i'll almost exclusively use common files and ifdef around
>    the subtleties -- but there is pretty much no difference between the
>    different families there at all afaict.
Agreed - I'd be most surprised if there were *any* commonality in code 
between different platforms. There should though be at least some within 
a particular family. I've done a fair bit on the STM32 now. Between 
different subfamilies some peripherals have identical code while others 
share quite a bit. I've pulled out the common stuff to a separate 
directory and included it where appropriate. The result is quite 
satisfying. I avoided ifdef mainly because the amount of commonality is 
quite large and this would make the code rather hard to read.
>
> * common peripherials
>
>    for more particular peripherials like usb, we should have a close look
>    whether we found common ancestry. arm producers might have bought ip
>    cores from third parties and not talk much about it. if we find
>    parallels, chances are it's really the same core and we can be done
>    with ifdef-ing again.
>
> * abstractions
>
>    arduino has its digitalRead and digitalWrite -- whenever such an
>    abstraction gets added, we should agree on what functionality can be
>    provided in general and for what it's not feasible.
That's where the unification would occur, but there again I wouldn't 
expect a great deal. Maybe just some basic set/reset/read/write. Haven't 
looked closely at USB yet but it seems focused on one STM32 subfamily as 
far as hardware drivers go. The rest is fairly generic.
>
> neither of that is what i'm aiming for with my generalizations, though
> -- i try to focus on what's common *because it's a cortex*. in that
> matter, i'm just overhauling my generalizations after i found out the
> irqs are already defined for most platforms.
>
> best regards
> chrysn
>
cheers, Ken