Luke Lau

> LGTM. Thanks. Will wait to hear back from @preames as well to make sure the correctness point is addressed

> Could this PR replace #78403 ? Not on its own, because the issue in #78403 stems from the fact that `DAGCombiner::foldSelectWithIdentityConstant` doesn't kick in with the sext/zext in between...

Sending out one last ping before I land this. I've rebased it and done another look over, it still seems correct to me. This PR allows the transform to happen...

> Everywhere you wrote "implicit" did you meant "implicit_def"? Yes, but I think nowadays this might actually be NoRegister

I think we also need to somehow make LNT Git SHA aware. At the moment if you want to the orders to make sense chronologically you manually need to specify...

> ~Change review status~ Hmm, It seems that it cannot be changed:( There's an option to dismiss stale reviews hidden away somewhere, I've accidentally done the same thing before :)...

> This is causing a failure with oggenc from llvm-test-suite for the rva23-evl config https://lab.llvm.org/staging/#/builders/210/builds/553 It looks like this involves first order recurrences, disabling vectorisation in the presence of VPFirstOrderRecurrencePHIRecipe...

FWIW I hope I'm not coming across as a perfectionist here, its just that I can't really follow the current code structure and so I can't really review it properly...

> Is it the case that unaligned scalar is supported while unaligned vector is not supported? #73971 cc @preames I don't remember the details, maybe t-head guys can help to...