[DAGCombine][X86][AArch64] 'A - (A & (B - 1))' -> 'A & (0 - B)' 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.

https://rise4fun.com/Alive/ZVdp

Name: ptr - (ptr & (alignment-1))  ->  ptr & (0 - alignment)
  %mask = add i64 %alignment, -1
  %bias = and i64 %ptr, %mask
  %r = sub i64 %ptr, %bias
=>
  %highbitmask = sub i64 0, %alignment
  %r = and i64 %ptr, %highbitmask

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

1 files changed, 15 insertions, 0 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 1415b1e37d1..5dbe29f5fbd 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -3104,6 +3104,21 @@ SDValue DAGCombiner::visitSUB(SDNode *N) {
                        DAG.getNode(ISD::SUB, DL, VT, N1.getOperand(1),
                                    N1.getOperand(0)));
 
+  // A - (A & (B - 1))  ->  A & (0 - B)
+  if (N1.getOpcode() == ISD::AND && N1.hasOneUse()) {
+    SDValue A = N1.getOperand(0);
+    SDValue BDec = N1.getOperand(1);
+    if (A != N0)
+      std::swap(A, BDec);
+    if (A == N0 && BDec.getOpcode() == ISD::ADD &&
+        isAllOnesOrAllOnesSplat(BDec->getOperand(1))) {
+      SDValue B = BDec.getOperand(0);
+      SDValue NegB =
+          DAG.getNode(ISD::SUB, DL, VT, DAG.getConstant(0, DL, VT), B);
+      return DAG.getNode(ISD::AND, DL, VT, A, NegB);
+    }
+  }
+
   // fold (X - (-Y * Z)) -> (X + (Y * Z))
   if (N1.getOpcode() == ISD::MUL && N1.hasOneUse()) {
     if (N1.getOperand(0).getOpcode() == ISD::SUB &&