[CostModel] avoid crashing while finding scalarization overhead
The constrained intrinsics have metadata arguments, so the tests here were crashing as noted in D90554 (and that was reverted even though this bug exists independently of that change).
Loading
Please sign in to comment