https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121740
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2025-09-02
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Ever confirmed|0 |1
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #3)
> (In reply to Richard Biener from comment #2)
> > p *ref
> > $9 = {ref = <component_ref 0x7ffff48513c0>, base = <var_decl
> > 0x7ffff4fe5ab0>,
> > offset = {coeffs = {320}}, size = {coeffs = {8}}, max_size = {coeffs =
> > {8}},
> > ref_alias_set = 21, base_alias_set = 0, volatile_p = false}
> >
> > so the load has alias-set zero and thus conflicts with SR.93_7->i. It has
> > alias-set zero because _Optional_payload_base has TYPE_TYPELESS_STORAGE.
>
> Though a store of a zero that does change a value of a zero.
Hmm, yes. So in principle there's code to handle that, but it only
opportunistically skips CONSTANT_CLASS_P rhs, not CTOR. We also only
skip stores of the same type (we're later using operand_equal_p), so
never aggregate copies. Fixable, I guess. Still it would require
quite some refactoring of the validity checking - you need to ensure
the = {} store will be not a partial def of the lookup (in this case
we're looking up a byte store, so "easy", but in a more general case).
It's the
/* If we reach a clobbering statement try to skip it and see if
we find a VN result with exactly the same value as the
possible clobber. In this case we can ignore the clobber
and return the found value. */
code in tree-ssa-sccvn.cc that does this.
I'll queue this.
