On 11/5/07, Rafael Espindola <[EMAIL PROTECTED]> wrote: > > If you're going to do this, why not just get rid of Subtarget > > altogether and just use BaseSubtarget? It can be set to 0 for targets > > that don't use it. Then again, it might not be a bad idea to use > > virtual getter/setters here. It already has virtual functions, and > > you're calling virtual functions through this pointer. So... :-) > > The problem with using only BaseSubtarget is that X86Subtarget adds a > lot of methods to the base implementation. Because of this the X86 > code really expects to see a pointer to the X86Subtarget and not > TargetSubtarget. I.E., a downcast is needed. > I was thinking the opposite way, but wrote down the other way. :-)
If you use just Subtarget in each target that implements it and then expose it through the getter, then you don't need the down-cast, only an upcast. So no RTTI, etc. That way there's no extra info for targets which don't have a subtarget, etc. What do you think? -bw _______________________________________________ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
