http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51920
Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
AssignedTo|unassigned at gcc dot |ebotcazou at gcc dot
|gnu.org |gnu.org
--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-01-29
23:23:19 UTC ---
> The output operand of all insns generating 'fpmerge' only accept the proper
> 'e'
> constraint.
'e' doesn't imply even-numbered though, there is no specific constraint for
that.
I think it's the long-standing issue with word-mode paradoxical subregs of FP
regs in 64-bit mode, for which both the PA and the SPARC ports have a kludge,
in the form of the CANNOT_CHANGE_MODE_CLASS macro.
I'm going to kludge again for this one by tweaking vector_init_fpmerge a bit.
