[Bug target/81863] [7 regression] -mword-relocations is unreliable

[yroux at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81863

--- Comment #5 from Yvan Roux <yroux at gcc dot gnu.org> ---
Thinking more about it, I think that the right place to fix it is in the define
of TARGET_HAVE_MOVT or TARGET_USE_MOVT, but I'm a bit confused with these two
macros.

My understanding of their semantic, is that it should be in TARGET_USE_MOVT
(the chip has movt support but we don't want to use it with when
-mword_relocations is used), but if we do so the issue movw/movt will be
generated anyway since arm_movt pattern and "j" contraint uses
TARGET_HAVE_MOVT.

In fact, it seems to me that most of TARGET_HAVE_MOVT should be replaced by
TARGET_USE_MOVT.  Kyrill, Thomas can you educate me ?