[Bug target/59405] Incorrect FP<->MMX transition during call/ret

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59405

--- Comment #8 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Yukhin Kirill from comment #5)
> I see. So, it seems like a limitation to passing vectors as arguments in
> 32-bit mode. We may implement something similar to `vzerroupper'
> autogeneration or simply close the bug as `user misunderstanding.'

We already tried this. Please see PR19161.

There is however, one problem - gcc warns about SSE vector argument without SSE
on 32bit targets. I will fix type_natural_mode.