https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125953
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Macleod from comment #1)
> Hmm. I would have said we did, but i don't see it.
>
> It would not be difficult to perform a global query if the stmt has no basic
> block set. I would have sworn we did that :-P
>
> Maybe we decided it doesn't make sense to pass in a stmt for context that
> isn't in the IL? there is no context in that case, so passing it in might
> be in error?
Yes, I definitely consider what the vectorizer does here an error. SCEV
will also blow up, and ranger might ask it?
> I'm OK wither way.. in range_of_expr, something like
>
> // If there is no statement, just get the global value.
> - if (!stmt)
> + if (!stmt || !gimple_bb (stmt))
but then we pass in 'stmt' as SSA_NAME_DEF_STMT (expr) (we should pass
in the context to query for, not the definition, right?), so this
wouldn't change anything. Ah, there it is:
// If there is no statement, just get the global value.
if (!stmt)
{
value_range tmp (TREE_TYPE (expr));
// If there is no global range for EXPR yet, try to evaluate it.
// This call sets R to a global range regardless.
if (!m_cache.get_global_range (r, expr))
{
gimple *s = SSA_NAME_DEF_STMT (expr);
// Calculate a range for S if it is safe to do so.
if (s && gimple_bb (s) && gimple_get_lhs (s) == expr)
^^^^^^^^^^^^^^^^^
return range_of_stmt (r, s);
and the safety in the if (stmt) case doesn't trigger:
else
{
basic_block bb = gimple_bb (stmt);
gimple *def_stmt = SSA_NAME_DEF_STMT (expr);
// If name is defined in this block, try to get an range from S.
if (def_stmt && gimple_bb (def_stmt) == bb)
because bb is NULL. I suppose to mimic the above safety check we need
to add && gimple_bb (def_stmt) here as well.
I'll play a bit with this and vectorizer fixes once I got a reduced testcase.
