Add missing CUDA kernel to tensor scan op

The TensorScanOp implementation was missing a CUDA kernel launch.
This adds a simple placeholder implementation.

1 files changed, 53 insertions, 65 deletions

diff --git a/unsupported/test/cxx11_tensor_scan.cpp b/unsupported/test/cxx11_tensor_scan.cpp
index bafa6c96e..af59aa3ef 100644
--- a/unsupported/test/cxx11_tensor_scan.cpp
+++ b/unsupported/test/cxx11_tensor_scan.cpp
@@ -14,87 +14,73 @@
 
 using Eigen::Tensor;
 
-template <int DataLayout, typename Type=float>
+template <int DataLayout, typename Type=float, bool Exclusive = false>
 static void test_1d_scan()
 {
-    int size = 50;
-    Tensor<Type, 1, DataLayout> tensor(size);
-    tensor.setRandom();
-    Tensor<Type, 1, DataLayout> result = tensor.cumsum(0);
-
-    VERIFY_IS_EQUAL(tensor.dimension(0), result.dimension(0));
-
-    float accum = 0;
-    for (int i = 0; i < size; i++) {
-      accum += tensor(i);
-      VERIFY_IS_EQUAL(result(i), accum);
-    }
-
-    accum = 1;
-    result = tensor.cumprod(0);
-    for (int i = 0; i < size; i++) {
-      accum *= tensor(i);
-      VERIFY_IS_EQUAL(result(i), accum);
-    }
-}
-
-template <int DataLayout, typename Type=float>
-static void test_1d_inclusive_scan()
-{
-    int size = 50;
-    Tensor<Type, 1, DataLayout> tensor(size);
-    tensor.setRandom();
-    Tensor<Type, 1, DataLayout> result = tensor.cumsum(0, true);
+  int size = 50;
+  Tensor<Type, 1, DataLayout> tensor(size);
+  tensor.setRandom();
+  Tensor<Type, 1, DataLayout> result = tensor.cumsum(0, Exclusive);
 
-    VERIFY_IS_EQUAL(tensor.dimension(0), result.dimension(0));
+  VERIFY_IS_EQUAL(tensor.dimension(0), result.dimension(0));
 
-    float accum = 0;
-    for (int i = 0; i < size; i++) {
+  float accum = 0;
+  for (int i = 0; i < size; i++) {
+    if (Exclusive) {
       VERIFY_IS_EQUAL(result(i), accum);
       accum += tensor(i);
+    } else {
+      accum += tensor(i);
+      VERIFY_IS_EQUAL(result(i), accum);
     }
+  }
 
-    accum = 1;
-    result = tensor.cumprod(0, true);
-    for (int i = 0; i < size; i++) {
+  accum = 1;
+  result = tensor.cumprod(0, Exclusive);
+  for (int i = 0; i < size; i++) {
+    if (Exclusive) {
       VERIFY_IS_EQUAL(result(i), accum);
       accum *= tensor(i);
+    } else {
+      accum *= tensor(i);
+      VERIFY_IS_EQUAL(result(i), accum);
     }
+  }
 }
 
 template <int DataLayout, typename Type=float>
 static void test_4d_scan()
 {
-    int size = 5;
-    Tensor<Type, 4, DataLayout> tensor(size, size, size, size);
-    tensor.setRandom();
+  int size = 5;
+  Tensor<Type, 4, DataLayout> tensor(size, size, size, size);
+  tensor.setRandom();
 
-    Tensor<Type, 4, DataLayout> result(size, size, size, size);
+  Tensor<Type, 4, DataLayout> result(size, size, size, size);
 
-    result = tensor.cumsum(0);
-    float accum = 0;
-    for (int i = 0; i < size; i++) {
-      accum += tensor(i, 0, 0, 0);
-      VERIFY_IS_EQUAL(result(i, 0, 0, 0), accum);
-    }
-    result = tensor.cumsum(1);
-    accum = 0;
-    for (int i = 0; i < size; i++) {
-      accum += tensor(0, i, 0, 0);
-      VERIFY_IS_EQUAL(result(0, i, 0, 0), accum);
-    }
-    result = tensor.cumsum(2);
-    accum = 0;
-    for (int i = 0; i < size; i++) {
-      accum += tensor(0, 0, i, 0);
-      VERIFY_IS_EQUAL(result(0, 0, i, 0), accum);
-    }
-    result = tensor.cumsum(3);
-    accum = 0;
-    for (int i = 0; i < size; i++) {
-      accum += tensor(0, 0, 0, i);
-      VERIFY_IS_EQUAL(result(0, 0, 0, i), accum);
-    }
+  result = tensor.cumsum(0);
+  float accum = 0;
+  for (int i = 0; i < size; i++) {
+    accum += tensor(i, 1, 2, 3);
+    VERIFY_IS_EQUAL(result(i, 1, 2, 3), accum);
+  }
+  result = tensor.cumsum(1);
+  accum = 0;
+  for (int i = 0; i < size; i++) {
+    accum += tensor(1, i, 2, 3);
+    VERIFY_IS_EQUAL(result(1, i, 2, 3), accum);
+  }
+  result = tensor.cumsum(2);
+  accum = 0;
+  for (int i = 0; i < size; i++) {
+    accum += tensor(1, 2, i, 3);
+    VERIFY_IS_EQUAL(result(1, 2, i, 3), accum);
+  }
+  result = tensor.cumsum(3);
+  accum = 0;
+  for (int i = 0; i < size; i++) {
+    accum += tensor(1, 2, 3, i);
+    VERIFY_IS_EQUAL(result(1, 2, 3, i), accum);
+  }
 }
 
 template <int DataLayout>
@@ -113,8 +99,10 @@ static void test_tensor_maps() {
 }
 
 void test_cxx11_tensor_scan() {
-  CALL_SUBTEST(test_1d_scan<ColMajor>());
-  CALL_SUBTEST(test_1d_scan<RowMajor>());
+  CALL_SUBTEST((test_1d_scan<ColMajor, float, true>()));
+  CALL_SUBTEST((test_1d_scan<ColMajor, float, false>()));
+  CALL_SUBTEST((test_1d_scan<RowMajor, float, true>()));
+  CALL_SUBTEST((test_1d_scan<RowMajor, float, false>()));
   CALL_SUBTEST(test_4d_scan<ColMajor>());
   CALL_SUBTEST(test_4d_scan<RowMajor>());
   CALL_SUBTEST(test_tensor_maps<ColMajor>());