diff --git a/src/OpTensorCreate.cpp b/src/OpTensorCreate.cpp
index 5bd7317d1..55dddb006 100644
--- a/src/OpTensorCreate.cpp
+++ b/src/OpTensorCreate.cpp
@@ -84,10 +84,7 @@ OpTensorCreate::postSubmit()
 {
     SPDLOG_DEBUG("Kompute OpTensorCreate postSubmit called");
 
-    // TODO: Remove and add a test that checks that the memory in 
-    // the staging tensor is actually storing the data
     SPDLOG_DEBUG("Kompute OpTensorCreate destroying staging tensors");
-    // TODO: This would cause issues if there is no CPU barrier
     this->mStagingTensors.clear();
 }
 
diff --git a/src/OpTensorSyncLocal.cpp b/src/OpTensorSyncLocal.cpp
index 37037ff99..8412c5063 100644
--- a/src/OpTensorSyncLocal.cpp
+++ b/src/OpTensorSyncLocal.cpp
@@ -89,7 +89,6 @@ OpTensorSyncLocal::postSubmit()
 
     // Remove all staging tensors as they are not required after operation
     SPDLOG_DEBUG("Kompute OpTensorSyncLocal destroying staging tensors");
-    // TODO: This would cause issues if there is no CPU barrier
     this->mStagingTensors.clear();
 }
 
diff --git a/test/TestOpTensorCopy.cpp b/test/TestOpTensorCopy.cpp
index 9325cacf5..5c44f3a39 100644
--- a/test/TestOpTensorCopy.cpp
+++ b/test/TestOpTensorCopy.cpp
@@ -27,6 +27,35 @@ TEST(TestOpTensorCopy, CopyDeviceToDeviceTensor) {
     EXPECT_EQ(tensorA->data(), tensorB->data());
 }
 
+TEST(TestOpTensorCopy, CopyDeviceToDeviceTensorMulti) {
+
+    kp::Manager mgr;
+
+    std::vector<float> testVecA{ 9, 8, 7 };
+    std::vector<float> testVecB{ 0, 0, 0 };
+    std::vector<float> testVecC{ 0, 0, 0 };
+
+    std::shared_ptr<kp::Tensor> tensorA{new kp::Tensor(testVecA)};
+    std::shared_ptr<kp::Tensor> tensorB{new kp::Tensor(testVecB)};
+    std::shared_ptr<kp::Tensor> tensorC{new kp::Tensor(testVecC)};
+
+    mgr.evalOpDefault<kp::OpTensorCreate>({tensorA, tensorB, tensorC});
+
+    EXPECT_TRUE(tensorA->isInit());
+    EXPECT_TRUE(tensorB->isInit());
+    EXPECT_TRUE(tensorC->isInit());
+
+    mgr.evalOpDefault<kp::OpTensorCopy>({tensorA, tensorB, tensorC});
+
+    EXPECT_EQ(tensorA->data(), tensorB->data());
+    EXPECT_EQ(tensorA->data(), tensorC->data());
+
+    // Making sure the GPU holds the same data
+    mgr.evalOpDefault<kp::OpTensorSyncLocal>({tensorB, tensorC});
+    EXPECT_EQ(tensorA->data(), tensorB->data());
+    EXPECT_EQ(tensorA->data(), tensorC->data());
+}
+
 TEST(TestOpTensorCopy, CopyDeviceToStagingTensor) {
 
     kp::Manager mgr;