[Bug middle-end/117932] [15 Regression] 100% compile-time slowdown from r15-518-g99b1daae18c095 for PR26854 testcase at -O2

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117932

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
So getting to sink heuristics, we might be disabling RTL if-conversion (cmov)
detection by sinking.  But disallowing sinking into (half-)diamonds doesn't
change anything so the testcase does not exhibit such cases.

I can't think of other good reasons to not sink into (empty) forwarders.