[llvm] [KnownBits] Add fast path for udiv by pow2 and overloads for shift ops by constant (PR #189779)

[Max Graey via llvm-commits]

MaxGraey wrote:

It looks [like this change](https://github.com/llvm/llvm-project/pull/189779/changes#diff-e6db17dae29e7ff730beec131ac8340805eea3252166d2fe57600ff0a251249cR790):
```diff

-      Known = KnownBits::shl(Known,
-                             KnownBits::makeConstant(APInt(BitWidth, ShiftAmt)),
+      Known = KnownBits::shl(Known, ShiftAmt,
                             /* NUW */ IOp->hasNoUnsignedWrap(),
                             /* NSW */ IOp->hasNoSignedWrap());
```

lead to regression for constant folding in [InstCombine/select-obo-peo-ops.ll](https://github.com/llvm/llvm-project/blob/main/llvm/test/Transforms/InstCombine/select-obo-peo-ops.ll#L140):
```diff
define i32 @test_shl_nuw_nsw__nsw_is_safe(i32 %x) {
+; CHECK-LABEL: @test_shl_nuw_nsw__nsw_is_safe(
+; CHECK-NEXT:    [[TMP1:%.*]] = or i32 [[X:%.*]], -83886080
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw i32 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = mul i32 [[TMP3]], [[TMP2]]
-; CHECK-NEXT:    ret i32 [[TMP4]]
 ;
  %1 = or i32 %x, -83886080
  %2 = icmp eq i32 %1, -83886079
```

```diff
define i32 @test_shl_nuw__nsw_is_safe(i32 %x) {
+; CHECK-LABEL: @test_shl_nuw__nsw_is_safe(
+; CHECK-NEXT:    [[TMP1:%.*]] = or i32 [[X:%.*]], -83886080
+; CHECK-NEXT:    [[TMP2:%.*]] = shl nsw i32 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[TMP2]], [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = mul i32 [[TMP3]], [[TMP2]]
-; CHECK-NEXT:    ret i32 [[TMP4]]
 ;
   %1 = or i32 %x, -83886080
   %2 = icmp eq i32 %1, -83886079
```

https://github.com/llvm/llvm-project/pull/189779