https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115881
Richard Sandiford <rsandifo at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |rsandifo at gcc dot
gnu.org
Status|NEW |ASSIGNED
--- Comment #3 from Richard Sandiford <rsandifo at gcc dot gnu.org> ---
Hmm, yeah. This is due to an inconsistency between simplify_subreg_regno
(which rightly refuses to generate (reg:SI sp)) and validate_subreg,
which instead uses subreg_offset_representable_p + some preparatory tests.
Those preparatory tests are similar, but not quite identical to, the first
tests in simplify_subre_regno. And they don't include the stack and frame
pointer tests.
In principle, it feels like the validate_subreg test should just be:
/* For hard registers, check that the subreg can be folded to a REG. */
if (reg && REG_P (reg) && HARD_REGISTER_P (reg))
return simplify_subreg_regno (REGNO (reg), imode, offset, omode) >= 0;
but I've no idea how much that will break. Will try it and see.
