In the ppc-expand-isel pass, we use stepForward() to update the liveness, this function is not recommended, because it need the accurate kill info.
In fact, we have the function computeAndAddLiveIns(), it will use stepBackward() to update the liveness. It's the recommended method to update the liveins.
I have use the option machine-verifier-strict-livein provided in https://reviews.llvm.org/D78586/ to test this patch.
In the patch D78586, there are 57 PowerPC cases failed, after using this patch, 54 cases can pass, and the left 3 cases should be the cases issue.
I notice computeAndAddLiveIns will call stepBackward, this comment says it will lead to cyclic dependence. Is it still the case for current code?