From: Jan W. <we...@ef...> - 2010-10-13 13:16:53
|
Value of AVR port of SDCC would be the significantly higher flexibility of SDCC than GCC, in more then one aspect. While avr-gcc one one side enjoys the massive investment of various companies into GCC, e.g. through the aggressive early optimisations, it also suffers from GCC being targeted primarily at "big" processors, e.g. through inefficiency in certain operations involving unnecessary integer promotions, lack of memory classes support, fixed pointer size (which does not fit all situations), to name just a few. It is basically impossible to get a "mere mortal's" request through to the developers; and given gcc's vast complexity it is hard to setup and maintain a homebrew fork and implement modifications into it. So, it is true, that given avr-gcc's maturity and support from Atmel (they have paid employees working on avr-gcc, avr-libc and related projects), it is the only good choice of open-source (at least formally) and no-cost C compiler for "serious work". However, I believe that AVR support in SDCC would provide valuable basis for experimentation. I played with the AVR backend a couple of hours, and while I see it has serious defficiencies, I am now able to compile a very simple C source (only assignments) to an avr-as (-gas) accepted form (it is still not linkable yet). I believe that with an effort of a couple of work-days (and work-nights ;-) ) I would be able to get a simple program through the whole chain. Okay, the usual problem: I don't have that time... Jan Waclawek |