On Wed, Jul 27, 2022 at 08:51:58AM +0100, Roger Sayle wrote: > > They can be, as clearly documented (and obvious from the code), but you > can > > not ever have that in the RTL stream, which is needed for your patch to do > > anything. > > That's the misunderstanding; neither this nor the previous SUBREG patch, > affect/change what is in the RTL stream, no COMPARE nodes are every > changed or modified, only eliminated by the propagation/fusion in combine > (or CSE).
There are no guarantees at all for that though? > We have --enable-checking=rtl to guarantee that the documented invariants > always hold in the RTL stream. That unfortunately only checks a few structural constraints, and not even all. For example not that STRICT_LOW_PART has a SUBREG as argument, which is documented, and the only thing that makes sense anyway. This is PR106101. Unfortunately many targets violate this. Segher
