Index: lib/Target/ARM/ARMTargetTransformInfo.cpp
===================================================================
--- lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -126,6 +126,13 @@
       return 0;
   }
 
+  // xor a, -1 can always be folded to MVN
+  if (Opcode == Instruction::Xor) {
+    int64_t NegImm = Imm.getSExtValue();
+    if (NegImm == -1)
+      return 0;
+  }
+
   return getIntImmCost(Imm, Ty);
 }
 
Index: test/CodeGen/Thumb/mvn.ll
===================================================================
--- /dev/null
+++ test/CodeGen/Thumb/mvn.ll
@@ -0,0 +1,136 @@
+; RUN: llc -mtriple=thumbv6m-eabi %s -o - | FileCheck %s
+
+; CHECK-LABEL: test8
+; CHECK: mvn
+define void @test8(i8* %a) {
+  %x = load i8, i8* %a
+  %xn = xor i8 %x, -1
+  store i8 %xn, i8* %a
+  ret void
+}
+
+; CHECK-LABEL: test8_2
+; CHECK: mvn
+; CHECK: mvn
+define void @test8_2(i8* %a, i8* %b) {
+  %x = load i8, i8* %a
+  %y = load i8, i8* %b
+  %xn = xor i8 %x, -1
+  %yn = xor i8 %y, -1
+  store i8 %xn, i8* %a
+  store i8 %yn, i8* %b
+  ret void
+}
+
+; CHECK-LABEL: loop8
+; CHECK: mvn
+define void @loop8(i8* %a) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %arrayidx = getelementptr inbounds i8, i8* %a, i32 %i
+  %x = load i8, i8* %arrayidx
+  %xn = xor i8 %x, -1
+  store i8 %xn, i8* %arrayidx
+  %inc = add nuw nsw i32 %i, 1
+  %exitcond = icmp eq i32 %inc, 10
+  br i1 %exitcond, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+
+; CHECK-LABEL: loop8_2
+; CHECK: mvn
+; CHECK: mvn
+define void @loop8_2(i8* %a, i8* %b) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %arrayidx = getelementptr inbounds i8, i8* %a, i32 %i
+  %arrayidx2 = getelementptr inbounds i8, i8* %b, i32 %i
+  %x = load i8, i8* %arrayidx
+  %y = load i8, i8* %arrayidx2
+  %xn = xor i8 %x, -1
+  %yn = xor i8 %y, -1
+  store i8 %xn, i8* %arrayidx
+  store i8 %yn, i8* %arrayidx2
+  %inc = add nuw nsw i32 %i, 1
+  %exitcond = icmp eq i32 %inc, 10
+  br i1 %exitcond, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+
+; CHECK-LABEL: test32
+; CHECK: mvn
+define void @test32(i32* %a) {
+  %x = load i32, i32* %a
+  %xn = xor i32 %x, -1
+  store i32 %xn, i32* %a
+  ret void
+}
+
+; CHECK-LABEL: test32_2
+; CHECK: mvn
+; CHECK: mvn
+define void @test32_2(i32* %a, i32* %b) {
+  %x = load i32, i32* %a
+  %y = load i32, i32* %b
+  %xn = xor i32 %x, -1
+  %yn = xor i32 %y, -1
+  store i32 %xn, i32* %a
+  store i32 %yn, i32* %b
+  ret void
+}
+
+; CHECK-LABEL: loop32
+; CHECK: mvn
+define void @loop32(i32* %a) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %arrayidx = getelementptr inbounds i32, i32* %a, i32 %i
+  %x = load i32, i32* %arrayidx
+  %xn = xor i32 %x, -1
+  store i32 %xn, i32* %arrayidx
+  %inc = add nuw nsw i32 %i, 1
+  %exitcond = icmp eq i32 %inc, 10
+  br i1 %exitcond, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+
+; CHECK-LABEL: loop32_2
+; CHECK: mvn
+; CHECK: mvn
+define void @loop32_2(i32* %a, i32* %b) {
+entry:
+  br label %for.body
+
+for.body:
+  %i = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %arrayidx = getelementptr inbounds i32, i32* %a, i32 %i
+  %arrayidx2 = getelementptr inbounds i32, i32* %b, i32 %i
+  %x = load i32, i32* %arrayidx
+  %y = load i32, i32* %arrayidx2
+  %xn = xor i32 %x, -1
+  %yn = xor i32 %y, -1
+  store i32 %xn, i32* %arrayidx
+  store i32 %yn, i32* %arrayidx2
+  %inc = add nuw nsw i32 %i, 1
+  %exitcond = icmp eq i32 %inc, 10
+  br i1 %exitcond, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+