[llvm] [InstCombine] Fold `X udiv Y` to `X lshr cttz(Y)` if Y is a power of 2 (PR #121386)

[via llvm-commits]

================
@@ -1632,6 +1632,16 @@ Instruction *InstCombinerImpl::visitUDiv(BinaryOperator &I) {
         I, Builder.CreateLShr(Op0, Res, I.getName(), I.isExact()));
   }
 
+  // Op0 udiv Op1 -> Op0 lshr cttz(Op1), if Op1 is a power of 2.
+  if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/ true, /*Depth*/ 0, &I)) {
+    // This will increase instruction count but it's okay
+    // since bitwise operations are substantially faster than
+    // division.
+    auto *Cttz =
+        Builder.CreateBinaryIntrinsic(Intrinsic::cttz, Op1, Builder.getTrue());
+    return BinaryOperator::CreateLShr(Op0, Cttz);
+  }
----------------
goldsteinn wrote:

You can preserve exact: https://alive2.llvm.org/ce/z/n8jnNn

Also it might be cleaner to merge this with the above fold using `takeLog2`. Maybe create a lambda like:

```
auto GetLog2Denom = [&](Value * Denom) {
  if (auto * R = takeLog2(Denom, ...) {
     return R;
   }
  if (IsKnownPowerOf2(Denom, ....) {
    return Builder.CreateBinaryIntrinsic(cttz, ....);
   }
   return nullptr;
}
```
Then:
```
if(auto * Res = GetLog2Denom(Op1)) {
 /// Above fold using `takeLog2`
}
```

https://github.com/llvm/llvm-project/pull/121386