> Yup, it was a bug. > I'm doing somewhat better down. > I've got the redundant store elimination portion working, with no > regressions. Congratulations! Hope to see it in gcc soon. Honza