> I think narrowing the loads and stores really needs to happen very late to > maximize the information about the semantic model of the program in the IR. > Now, this is predicated on us being *able* to do that narrowing. It was very > easy in x86 (like, one tiny bug to fix, almost all of it "just worked") so > I'm optimistic this is actually easy in general. Let me know if not though.
How late? Would a late IR pass be OK maybe? Cheers Rafael _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
