[PATCH] Improve CodeGenDAGPattern support with multiple explicit results in an instruction pattern

[hfinkel at anl.gov]

In http://reviews.llvm.org/D8073#134733, @craig.topper wrote:

> To be honest I don't know if I know the code well enough to know if that
>  can fail. Just seemed bad to blindly assume there was only one type there.
>  There are quite a few such "unhandled" asserts in this file so I just wrote
>  the same.


Okay, fair enough. You're right, the asserts on this condition are like this for some reason. Let's keep it this way in this patch, and we'll improve them all together later.


http://reviews.llvm.org/D8073

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/