https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124023
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Depends on| |115082
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #1)
> I have not looked if there is a pattern to do that or not (without the
> second use of captured+1.
There is not a pattern that handles:
```
bool f(int captured)
{
return (((captured)+1)&1) == 0;
}
```
what does it is VRP.
VRP does it is by folding `zero_one_valued == 0` into `(bool)(zero_one_valued ^
1)` And then there are patterns to handle that (in some award way)
VRP does not change GIMPLE_COND here since == 0 is simplier than the extra xor.
Note `(((captured)+1)&1)` can be canonicalize as `1 & ~a` which is PR 115082.
and then `(1&~a) == 0` is not currently simplified into `a != 0` though.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115082
[Bug 115082] Canonical form for `(a+oddCST)&1` or `1 & ~a`
