diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 5c14ed6..a76531a 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -2578,9 +2578,9 @@
   switch (IntNo) {
   default: return SDValue();    // Don't custom lower most intrinsics.
   case Intrinsic::arm_rbit: {
-    assert(Op.getOperand(0).getValueType() == MVT::i32 &&
+    assert(Op.getOperand(1).getValueType() == MVT::i32 &&
            "RBIT intrinsic must have i32 type!");
-    return DAG.getNode(ARMISD::RBIT, dl, MVT::i32, Op.getOperand(0));
+    return DAG.getNode(ARMISD::RBIT, dl, MVT::i32, Op.getOperand(1));
   }
   case Intrinsic::arm_thread_pointer: {
     EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy();
diff --git a/llvm/test/CodeGen/AArch64/rbit.ll b/llvm/test/CodeGen/AArch64/rbit.ll
new file mode 100644
index 0000000..3404ae4
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/rbit.ll
@@ -0,0 +1,20 @@
+; RUN: llc -mtriple=aarch64-eabi %s -o - | FileCheck %s
+
+; CHECK-LABEL: rbit32
+; CHECK: rbit w0, w0
+define i32 @rbit32(i32 %t) {
+entry:
+  %rbit.i = call i32 @llvm.aarch64.rbit.i32(i32 %t)
+  ret i32 %rbit.i
+}
+
+; CHECK-LABEL: rbit64
+; CHECK: rbit x0, x0
+define i64 @rbit64(i64 %t) {
+entry:
+  %rbit.i = call i64 @llvm.aarch64.rbit.i64(i64 %t)
+  ret i64 %rbit.i
+}
+
+declare i64 @llvm.aarch64.rbit.i64(i64)
+declare i32 @llvm.aarch64.rbit.i32(i32)
diff --git a/llvm/test/CodeGen/ARM/rbit.ll b/llvm/test/CodeGen/ARM/rbit.ll
new file mode 100644
index 0000000..41f866f
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/rbit.ll
@@ -0,0 +1,20 @@
+; RUN: llc -mtriple=armv8-eabi %s -o - | FileCheck %s
+
+; CHECK-LABEL: rbit
+; CHECK: rbit r0, r0
+define i32 @rbit(i32 %t) {
+entry:
+  %rbit = call i32 @llvm.arm.rbit(i32 %t)
+  ret i32 %rbit
+}
+
+; CHECK-LABEL: rbit_constant
+; CHECK: mov r0, #0
+; CHECK: rbit r0, r0
+define i32 @rbit_constant() {
+entry:
+  %rbit.i = call i32 @llvm.arm.rbit(i32 0)
+  ret i32 %rbit.i
+}
+
+declare i32 @llvm.arm.rbit(i32)