[Bug debug/54953] [4.8 Regression] New sra-1.c FAILs on powerpc

["aoliva at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54953

--- Comment #2 from Alexandre Oliva <aoliva at gcc dot gnu.org> 2012-10-26 07:42:07 UTC ---
Without your poposed change, AFTER_WITH_REG is only used while adding
REG_UNUSED marks to REG defs.  That the REG def is unused means it's going to
be discarded and, when it is, either the debug temp will be reset altogether,
or we'll propagate the expression stored in the REG to uses of the REG (e.g.,
the debug use, if we omit the debug temp, or the debug temp, if we needlessly
add it), or we (should?) emit another debug temp BEFORE_WITH_VALUE and use that
instead of the REG (in the orginal debug use or in the needless debug temp). 
In any of these cases, the debug temp is useless.

All that siad, it's probably not a terribly important optimization, so if the
above wouldn't make sense in a comment before the questionable hunk, we might
as well drop it.  Adding REG_FORCE for this new use is probably fine too.