[DAGCombine][X86][Thumb2/LowOverheadLoops] `A - (A & C)` -> `A & (~C)` fold (PR44448)

While we do manage to fold integer-typed IR in middle-end,
we can't do that for the main motivational case of pointers.

There is @llvm.ptrmask() intrinsic which may or may not be helpful,
but i'm not sure it is fully considered canonical yet,
not everything is fully aware of it likely.

Name: PR44448  ptr - (ptr & C) -> ptr & (~C)
%bias = and i32 %ptr, C
%r = sub i32 %ptr, %bias
  =>
%r = and i32 %ptr, ~C

See
  https://bugs.llvm.org/show_bug.cgi?id=44448
  https://reviews.llvm.org/D71499

1 files changed, 10 insertions, 0 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index cd6d33ffa9e..dad47944a0f 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -3119,6 +3119,16 @@ SDValue DAGCombiner::visitSUB(SDNode *N) {
     }
   }
 
+  // A - (A & C) -> A & (~C)
+  if (N1.getOpcode() == ISD::AND && N1.getOperand(0) == N0 &&
+      isConstantOrConstantVector(N1.getOperand(1), /*NoOpaques=*/true)) {
+    SDValue InvC =
+        DAG.FoldConstantArithmetic(ISD::XOR, DL, VT, N1.getOperand(1).getNode(),
+                                   DAG.getAllOnesConstant(DL, VT).getNode());
+    assert(InvC && "Constant folding failed");
+    return DAG.getNode(ISD::AND, DL, VT, N0, InvC);
+  }
+
   // fold (X - (-Y * Z)) -> (X + (Y * Z))
   if (N1.getOpcode() == ISD::MUL && N1.hasOneUse()) {
     if (N1.getOperand(0).getOpcode() == ISD::SUB &&