http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54739
--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-10-04 07:58:31 UTC --- Thanks. The code generated for the testcase is certainly better on the SPARC if lower-subreg doesn't split the DImode objects (which probably means that the patch was also beneficial for this architecture) so I'll go the ARM way.