http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58442 --- Comment #2 from Jan-Benedict Glaw <jbglaw@lug-owl.de> --- So r0 is waay off. As we're far into the function (950) and fiddling with r0, I guess this is the final part, preparing to return from here. An assembler dump of the whole function would be helpful I hope.