[llvm] [SimplifyCFG] Add optimization for switches of powers of two (PR #70977)

[via llvm-commits]

================
@@ -6839,6 +6836,74 @@ static bool ReduceSwitchRange(SwitchInst *SI, IRBuilder<> &Builder,
   return true;
 }
 
+static bool simplifySwitchOfPowersOfTwo(SwitchInst *SI, IRBuilder<> &Builder,
+                                        const DataLayout &DL,
+                                        const TargetTransformInfo &TTI) {
+
+  Value *Condition = SI->getCondition();
+  LLVMContext &Context = SI->getContext();
+  auto *CondTy = cast<IntegerType>(Condition->getType());
+
+  if (CondTy->getIntegerBitWidth() > 64 ||
+      !DL.fitsInLegalInteger(CondTy->getIntegerBitWidth()))
+    return false;
+
+  const auto CttzIntrinsicCost = TTI.getIntrinsicInstrCost(
+      IntrinsicCostAttributes(Intrinsic::cttz, CondTy,
+                              {Condition, ConstantInt::getTrue(Context)}),
+      TTI::TCK_SizeAndLatency);
+
+  if (CttzIntrinsicCost > TTI::TCC_Basic) {
----------------
DKay7 wrote:

Could you please help me with this? Should I just find an arch with a slow cttz, then specify it with `-mtriple` and check that the transformation won't be applied, or should I do something smarter? Thank you.

https://github.com/llvm/llvm-project/pull/70977