On Thu, 2013-06-13 at 10:49 +0100, Florian Fainelli wrote: > We are in the slow process to switch to Device Tree to precisely > eliminate all of this (although not everyone agrees yet on the > details). Hopefully you should not see such things in the future. I don't see how DT help here (hint: it never does) if the driver knows at compile time how registers are shuffled, it can remove the indirection if you use runtime cpu detection or DT, it cannot in fact, this patch already adds another layer of indirection with that "dma_chan_width" thing that the compiler has no way to optimize out, defeating the purpose of single SOC optimization. so we might as well force multi SOC support -- Maxime
