LGTM. Thank for adding the test.

In D85599#2206033, @efriedma wrote:

Fixing a MachineVerifier issue by patching clang is generally wrong; if the IR is valid, the backend should do something sane with it. If the IR isn't valid, the IR Verifier should complain.

Here, I'd say the IR is valid; adding a special case to the IR rules here would be a pain. So the right fix involves changing isel and/or the definitions of the the instructions in question.

Remoeve isTerminator for TRAP instruction.

Fix the case.

Add a new func setKillAfterFolding.

In D84927#2185618, @MaskRay wrote:

If your build is configured with -DLLVM_ENABLE_EXPENSIVE_CHECKS=on, verifyer passes may be added even without -verify-machineinstrs. I assume you have tested them...

Remove the CR1

FIx the comments.

Update the patch to follow reviewers' comments.

Simplify the code.

Update the patch to follow reviewers' comments.

Remove the hasP8Vector condition.

The new test makes more sense to me. Can you please use this one ? So, yes, mark the v1i128 as expand so that, we won't have any setcc with v1i128 after type legalization, which fix the crashing you see when we are trying to select an opcode for v1i128.

Update the test case.

In D84238#2165944, @steven.zhang wrote:

I still think there is something missing here. What you set the legalize action for SETCC is under P8 altivec, but why the case passed for P7 ? And this case passed without your patch :

P7 will convert result from vector v1i128 setcc to two scalar setcc i64.

Remove le for PWR7.

In D84238#2165867, @steven.zhang wrote:

Your test didn't crash even without your fix.

Add more test cases.

Don't call fixupImplicitDefUseOperands ().

Add more cases.

Add the case for MTFSF.

Remove the dump code.

Add IsNeedFixImplicitOps to call the fixupImplicitDefUseOperands when needed.

ping. @thegameg @qcolombet , do you have any comments for this patch?

Merge https://reviews.llvm.org/D83471 into this patch.

Rebase.

In D81360#2159250, @jsji wrote:

Also why this depends on D76042?

Call IsKilledFor() to fix the flag.

Fix the error case.

ping...

Add CHECK for the case.

@dexonsmith Do you have any comments for this patch?

Fix XSRDPI/XSRDPIM/XSRDPIP/XSRDPIZ, XVRDPI/XVRDPIM XVRDPIP XSRDPIZ, XVRSPI/XVRSPIM XVRSPIP XSRSPIZ

Add a test case.

Rebase.

Add guard for OldKill.

No concrete use case, abandoned it.

@bjope Do you have any other comments for this patch?

In D80077#2045630, @lkail wrote:

It looks like a bug in phi-node-elimination to me. When critical edge bb.0 -> bb.2 is splitted and bb.3 is inserted, undef %undef is transformed into %3 = IMPLICIT_DEF in bb.3. LV in bb.2 is not updated in memory(If we run phi-node-elimination and twoaddressinstruction separately, we won't run into crash).