[PATCH] Teach the inliner how to preserve musttail invariants

[Chandler Carruth]

Ok, this makes more sense now. Thahnks for the clue-bat.

I want to ruminate on the bit about not needing stackrestore. I think I believe you, but I don't fully understand why yet and I would like to. But below are some easy comments to just clean up code.

================
Comment at: lib/Transforms/Utils/InlineFunction.cpp:697
@@ +696,3 @@
+      for (Instruction &I : *BB) {
+        if (CallInst *CI = dyn_cast<CallInst>(&I)) {
+          CallInst::TailCallKind ChildTCK = CI->getTailCallKind();
----------------
nit: invert and use continue to reduce indentation

================
Comment at: lib/Transforms/Utils/InlineFunction.cpp:794
@@ +793,3 @@
+
+      if (NeedBitCast) {
+        // Strip any possible inlined bitcast and grab the musttail call.
----------------
nit: as above, invert and continue to reduce indentation.

================
Comment at: lib/Transforms/Utils/InlineFunction.cpp:795-800
@@ +794,8 @@
+      if (NeedBitCast) {
+        // Strip any possible inlined bitcast and grab the musttail call.
+        Value *RV = RI->getReturnValue();
+        BitCastInst *OldCast = dyn_cast<BitCastInst>(RV);
+        if (OldCast)
+          RV = OldCast->getOperand(0);
+        assert(cast<CallInst>(RV)->isMustTailCall());
+
----------------
Rather than duplicating isPrecededByMustTailCall, make that function getPrecedingMustTailCallInst or some such, and have it return null when there is no call? That should make this loop simplify quite a bit.

http://reviews.llvm.org/D3491