[Bug target/106736] [13 Regression] ICE in gen_movxo, at config/rs6000/mma.md:333

[avinashd at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106736

Avinash Jayakar <avinashd at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |avinashd at gcc dot gnu.org

--- Comment #22 from Avinash Jayakar <avinashd at gcc dot gnu.org> ---
Would it make sense to not make the type "__vector_pair" and "__vector_quad" be
specific to MMA, and instead be specific to VSX? 
So if the target does not have MMA support, the mov operation can be done using
2 or more loads/stores, which should eliminate the ICEs that were seen with
this and linked bugs.
This is because this type could be potentially be used for future features that
may not be related to MMA.
Any suggestions?