http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53592
Richard Guenther <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2012-06-11
CC| |rguenth at gcc dot gnu.org
Ever Confirmed|0 |1
--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-11
10:15:22 UTC ---
The gimplifier marks 'result' as DECL_GIMPLE_REG_P here:
/* Gimplify a BIND_EXPR. Just voidify and recurse. */
static enum gimplify_status
gimplify_bind_expr (tree *expr_p, gimple_seq *pre_p)
{
...
/* Preliminarily mark non-addressed complex variables as eligible
for promotion to gimple registers. We'll transform their uses
as we find them. */
if ((TREE_CODE (TREE_TYPE (t)) == COMPLEX_TYPE
|| TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
&& !TREE_THIS_VOLATILE (t)
&& (TREE_CODE (t) == VAR_DECL && !DECL_HARD_REGISTER (t))
&& !needs_to_live_in_memory (t))
DECL_GIMPLE_REG_P (t) = 1;
which means it expects the Frontends to have marked 'result' as
TREE_ADDRESSABLE
if they assign to it in this way. GENERIC from the frontend looks like
Ssa_View.Test_View_Ssa_Lhs (const ssa_view__fpt x0, const ssa_view__fpt x1)
{
ssa_view__vec result;
ssa_view__vec result;
VIEW_CONVERT_EXPR<ssa_view__fpt[2]>(result)[0] = (ssa_view__fpt) x0;
return result;
}
another possibility is, as the comment explains, that this is supposed to
be done for COMPLEX_TYPE only, as only component-accesses to those are
"transformed" properly. But note that an access like the above, via
a VIEW_CONVERTed type are not necessarily handled.
Well. Confirmed.
