On Mon, May 7, 2012 at 9:14 PM, John David Anglin <dave.ang...@bell.net> wrote:
> The maintenance problems that occur are usually not related to the code > generation. They largely occur because the target is big endian, strict > alignment, callee copies, etc. The delayed branch support has caused > many subtle bugs. For this reason, I don't think simplifying the backend > will help much. Here I disagree. IMHO simplifying any back end is a win, because it lowers the threshold to "modernize it". For instance for HP-PA, I would like to play with Bernd's new delay branch schedule model (http://gcc.gnu.org/ml/gcc-patches/2011-05/msg00753.html) and that will simply be a less daunting job if the back end is smaller. It will also be easier to test. Ciao! Steven
