[WIP][SimplifyCFG] Adjust sinking strategy for conditional predecessors.

[xbolva00]

llvm/lib/Transforms/Utils/SimplifyCFG.cpp

2258

Please add extra comment to explain your change.

[DianQK]

I will write the new idea in D155395.

llvm/lib/Transforms/Utils/SimplifyCFG.cpp

2258

This is a trick change in WIP.
I want to find a more suitable sink strategy to avoid missing out on some good opportunities.
I have reconsidered the optimization approach for D155395, and sink may not be the appropriate route.

llvm/lib/Transforms/Utils/SimplifyCFG.cpp

if (!followedByDeoptOrUnreachable) {
     // If no instructions can be sunk, early-return.
     if (ScanIdx == 0)
       return false;
   }
 
   bool Changed = false;
 
   if (HaveNonUnconditionalPredecessors) {
-    if (!followedByDeoptOrUnreachable) {
+    if (!followedByDeoptOrUnreachable && ScanIdx < 2) {
       // It is always legal to sink common instructions from unconditional
       // predecessors. However, if not all predecessors are unconditional,
       // this transformation might be pessimizing. So as a rule of thumb,
       // don't do it unless we'd sink at least one non-speculatable instruction.
       // See https://bugs.llvm.org/show_bug.cgi?id=30244
       LRI.reset();
       int Idx = 0;
       bool Profitable = false;
       while (Idx < ScanIdx) {
         if (!isSafeToSpeculativelyExecute((*LRI)[0])) {
           Profitable = true;
           break;