Here's a f'example.
I've been working on a framework that is basically a HAL like the Arduino (and does implement the Arduino functions) but has a lot more, such as string parsing, FIFOs etc.
Last I looked it was about 20k in size, a bit large for a Uno but no problems for a chip with a lot of flash. Now of course you don't have to go to an ARM to get a lot of flash, but with the smaller ARMs available for < $2 I just reckon it's easier to use them. Hopefully I can drop this code onto the Due when it arrives but it's not worth porting it to a Uno.
Rob