... you need to port your '51 product with as little as possible pain. Now we don't know if you have some code space restrictions on the AVR board, but my first guess would be no.

If this is the case, the questions you might want to ask in relation to gcc would be:

1. how easy is to migrate from '51/Keil to AVR/gcc, maybe in comparison with the other AVR compilers;

2. whether you can get (probably paid) support with some guarantee on gcc/avr in case of troubles (this might sound strange but I'd not say impossible).

There is no doubt on the (monetary) price of the compiler itself :-). Of course you might have your doubts regarding the quality (in terms of bugs) of the present state of the compiler, too. Now the sheer quantity of the compiler's users give at least a wide tester's base, which is certainly a desirable thing, regardless of if they are professionals or not; although the pressure to fix the bugs is certainly less than at a paid product.


JW