Extend support for Packet16b:

* Add ptranspose<*,4> to support matmul and add unit test for Matrix<bool> * Matrix<bool>
* work around a bug in slicing of Tensor<bool>.
* Add tensor tests

This speeds up matmul for boolean matrices by about 10x

name                            old time/op             new time/op             delta
BM_MatMul<bool>/8                267ns ± 0%              479ns ± 0%  +79.25%          (p=0.008 n=5+5)
BM_MatMul<bool>/32              6.42µs ± 0%             0.87µs ± 0%  -86.50%          (p=0.008 n=5+5)
BM_MatMul<bool>/64              43.3µs ± 0%              5.9µs ± 0%  -86.42%          (p=0.008 n=5+5)
BM_MatMul<bool>/128              315µs ± 0%               44µs ± 0%  -85.98%          (p=0.008 n=5+5)
BM_MatMul<bool>/256             2.41ms ± 0%             0.34ms ± 0%  -85.68%          (p=0.008 n=5+5)
BM_MatMul<bool>/512             18.8ms ± 0%              2.7ms ± 0%  -85.53%          (p=0.008 n=5+5)
BM_MatMul<bool>/1k               149ms ± 0%               22ms ± 0%  -85.40%          (p=0.008 n=5+5)

6 files changed, 180 insertions, 139 deletions

diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorMorphing.h b/unsupported/Eigen/CXX11/src/Tensor/TensorMorphing.h
index d04dcae17..f107d1b19 100644
--- a/unsupported/Eigen/CXX11/src/Tensor/TensorMorphing.h
+++ b/unsupported/Eigen/CXX11/src/Tensor/TensorMorphing.h
@@ -456,7 +456,9 @@ struct TensorEvaluator<const TensorSlicingOp<StartIndices, Sizes, ArgType>, Devi
     // slice offsets and sizes.
     IsAligned         = false,
     PacketAccess      = TensorEvaluator<ArgType, Device>::PacketAccess,
-    BlockAccess       = TensorEvaluator<ArgType, Device>::BlockAccess,
+    BlockAccess       = TensorEvaluator<ArgType, Device>::BlockAccess &&
+                        // FIXME: Temporary workaround for bug in slicing of bool tensors.
+                        !internal::is_same<typename internal::remove_const<Scalar>::type, bool>::value,
     PreferBlockAccess = true,
     Layout            = TensorEvaluator<ArgType, Device>::Layout,
     CoordAccess       = false,
@@ -525,7 +527,6 @@ struct TensorEvaluator<const TensorSlicingOp<StartIndices, Sizes, ArgType>, Devi
 
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; }
 
-
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(EvaluatorPointerType data) {
     m_impl.evalSubExprsIfNeeded(NULL);
     if (!NumTraits<typename internal::remove_const<Scalar>::type>::RequireInitialization
@@ -559,14 +560,14 @@ struct TensorEvaluator<const TensorSlicingOp<StartIndices, Sizes, ArgType>, Devi
     }
     return true;
   }
-  
+
 #ifdef EIGEN_USE_THREADS
   template <typename EvalSubExprsCallback>
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void evalSubExprsIfNeededAsync(
       EvaluatorPointerType data, EvalSubExprsCallback done) {
     m_impl.evalSubExprsIfNeededAsync(nullptr, [done](bool) { done(true); });
   }
-#endif  // EIGEN_USE_THREADS  
+#endif  // EIGEN_USE_THREADS
 
   EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void cleanup() {
     m_impl.cleanup();
diff --git a/unsupported/test/cxx11_tensor_block_eval.cpp b/unsupported/test/cxx11_tensor_block_eval.cpp
index 226c495aa..a7a49fa1f 100644
--- a/unsupported/test/cxx11_tensor_block_eval.cpp
+++ b/unsupported/test/cxx11_tensor_block_eval.cpp
@@ -233,7 +233,7 @@ static void test_eval_tensor_binary_expr_block() {
   rhs.setRandom();
 
   VerifyBlockEvaluator<T, NumDims, Layout>(
-      lhs + rhs, [&dims]() { return RandomBlock<Layout>(dims, 1, 10); });
+      lhs * rhs, [&dims]() { return RandomBlock<Layout>(dims, 1, 10); });
 }
 
 template <typename T, int NumDims, int Layout>
@@ -274,7 +274,7 @@ static void test_eval_tensor_broadcast() {
   // Check that desc.destination() memory is not shared between two broadcast
   // materializations.
   VerifyBlockEvaluator<T, NumDims, Layout>(
-      input.broadcast(bcast) + input.square().broadcast(bcast),
+      input.broadcast(bcast) * input.square().broadcast(bcast),
       [&bcasted_dims]() { return SkewedInnerBlock<Layout>(bcasted_dims); });
 }
 
@@ -509,7 +509,7 @@ static void test_eval_tensor_reshape_with_bcast() {
   DSizes<Index, 2> dims(dim, dim);
 
   VerifyBlockEvaluator<T, 2, Layout>(
-      lhs.reshape(reshapeLhs).broadcast(bcastLhs) +
+      lhs.reshape(reshapeLhs).broadcast(bcastLhs) *
           rhs.reshape(reshapeRhs).broadcast(bcastRhs),
       [dims]() { return SkewedInnerBlock<Layout, 2>(dims); });
 }
@@ -529,11 +529,11 @@ static void test_eval_tensor_forced_eval() {
   DSizes<Index, 2> dims(dim, dim);
 
   VerifyBlockEvaluator<T, 2, Layout>(
-      (lhs.broadcast(bcastLhs) + rhs.broadcast(bcastRhs)).eval().reshape(dims),
+      (lhs.broadcast(bcastLhs) * rhs.broadcast(bcastRhs)).eval().reshape(dims),
       [dims]() { return SkewedInnerBlock<Layout, 2>(dims); });
 
   VerifyBlockEvaluator<T, 2, Layout>(
-      (lhs.broadcast(bcastLhs) + rhs.broadcast(bcastRhs)).eval().reshape(dims),
+      (lhs.broadcast(bcastLhs) * rhs.broadcast(bcastRhs)).eval().reshape(dims),
       [dims]() { return RandomBlock<Layout, 2>(dims, 1, 50); });
 }
 
@@ -755,7 +755,39 @@ static void test_assign_to_tensor_shuffle() {
 #define CALL_SUBTEST_PART(PART) \
   CALL_SUBTEST_##PART
 
-#define CALL_SUBTESTS_DIMS_LAYOUTS(PART, NAME)           \
+#define CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(PART, NAME)           \
+  CALL_SUBTEST_PART(PART)((NAME<float, 1, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 2, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 3, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 4, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 5, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 1, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 2, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, 5, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 1, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 2, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 3, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 4, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 5, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 1, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 2, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<int, 5, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 1, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 2, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 3, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 4, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 5, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 1, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 2, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 4, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, 5, ColMajor>()))
+
+#define CALL_SUBTESTS_DIMS_LAYOUTS(PART, NAME)     \
   CALL_SUBTEST_PART(PART)((NAME<float, 1, RowMajor>())); \
   CALL_SUBTEST_PART(PART)((NAME<float, 2, RowMajor>())); \
   CALL_SUBTEST_PART(PART)((NAME<float, 3, RowMajor>())); \
@@ -767,36 +799,38 @@ static void test_assign_to_tensor_shuffle() {
   CALL_SUBTEST_PART(PART)((NAME<float, 4, ColMajor>())); \
   CALL_SUBTEST_PART(PART)((NAME<float, 5, ColMajor>()))
 
-#define CALL_SUBTESTS_LAYOUTS(PART, NAME)             \
+#define CALL_SUBTESTS_LAYOUTS_TYPES(PART, NAME)       \
   CALL_SUBTEST_PART(PART)((NAME<float, RowMajor>())); \
-  CALL_SUBTEST_PART(PART)((NAME<float, ColMajor>()))
+  CALL_SUBTEST_PART(PART)((NAME<float, ColMajor>()));  \
+  CALL_SUBTEST_PART(PART)((NAME<bool, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, ColMajor>()))
 
 EIGEN_DECLARE_TEST(cxx11_tensor_block_eval) {
   // clang-format off
-  CALL_SUBTESTS_DIMS_LAYOUTS(1, test_eval_tensor_block);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(1, test_eval_tensor_block);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(1, test_eval_tensor_binary_expr_block);
   CALL_SUBTESTS_DIMS_LAYOUTS(1, test_eval_tensor_unary_expr_block);
-  CALL_SUBTESTS_DIMS_LAYOUTS(1, test_eval_tensor_binary_expr_block);
   CALL_SUBTESTS_DIMS_LAYOUTS(2, test_eval_tensor_binary_with_unary_expr_block);
-  CALL_SUBTESTS_DIMS_LAYOUTS(2, test_eval_tensor_broadcast);
-  CALL_SUBTESTS_DIMS_LAYOUTS(2, test_eval_tensor_reshape);
-  CALL_SUBTESTS_DIMS_LAYOUTS(3, test_eval_tensor_cast);
-  CALL_SUBTESTS_DIMS_LAYOUTS(3, test_eval_tensor_select);
-  CALL_SUBTESTS_DIMS_LAYOUTS(3, test_eval_tensor_padding);
-  CALL_SUBTESTS_DIMS_LAYOUTS(4, test_eval_tensor_chipping);
-  CALL_SUBTESTS_DIMS_LAYOUTS(4, test_eval_tensor_generator);
-  CALL_SUBTESTS_DIMS_LAYOUTS(4, test_eval_tensor_reverse);
-  CALL_SUBTESTS_DIMS_LAYOUTS(5, test_eval_tensor_slice);
-  CALL_SUBTESTS_DIMS_LAYOUTS(5, test_eval_tensor_shuffle);
-
-  CALL_SUBTESTS_LAYOUTS(6, test_eval_tensor_reshape_with_bcast);
-  CALL_SUBTESTS_LAYOUTS(6, test_eval_tensor_forced_eval);
-  CALL_SUBTESTS_LAYOUTS(6, test_eval_tensor_chipping_of_bcast);
-
-  CALL_SUBTESTS_DIMS_LAYOUTS(7, test_assign_to_tensor);
-  CALL_SUBTESTS_DIMS_LAYOUTS(7, test_assign_to_tensor_reshape);
-  CALL_SUBTESTS_DIMS_LAYOUTS(7, test_assign_to_tensor_chipping);
-  CALL_SUBTESTS_DIMS_LAYOUTS(8, test_assign_to_tensor_slice);
-  CALL_SUBTESTS_DIMS_LAYOUTS(8, test_assign_to_tensor_shuffle);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(2, test_eval_tensor_broadcast);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(2, test_eval_tensor_reshape);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(3, test_eval_tensor_cast);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(3, test_eval_tensor_select);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(3, test_eval_tensor_padding);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(4, test_eval_tensor_chipping);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(4, test_eval_tensor_generator);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(4, test_eval_tensor_reverse);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(5, test_eval_tensor_slice);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(5, test_eval_tensor_shuffle);
+
+  CALL_SUBTESTS_LAYOUTS_TYPES(6, test_eval_tensor_reshape_with_bcast);
+  CALL_SUBTESTS_LAYOUTS_TYPES(6, test_eval_tensor_forced_eval);
+  CALL_SUBTESTS_LAYOUTS_TYPES(6, test_eval_tensor_chipping_of_bcast);
+
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(7, test_assign_to_tensor);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(7, test_assign_to_tensor_reshape);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(7, test_assign_to_tensor_chipping);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(8, test_assign_to_tensor_slice);
+  CALL_SUBTESTS_DIMS_LAYOUTS_TYPES(8, test_assign_to_tensor_shuffle);
 
   // Force CMake to split this test.
   // EIGEN_SUFFIXES;1;2;3;4;5;6;7;8
diff --git a/unsupported/test/cxx11_tensor_block_io.cpp b/unsupported/test/cxx11_tensor_block_io.cpp
index a6f7a44cd..52f7dde9b 100644
--- a/unsupported/test/cxx11_tensor_block_io.cpp
+++ b/unsupported/test/cxx11_tensor_block_io.cpp
@@ -415,7 +415,15 @@ static void test_block_io_squeeze_ones() {
   CALL_SUBTEST((NAME<float, 1, ColMajor>())); \
   CALL_SUBTEST((NAME<float, 2, ColMajor>())); \
   CALL_SUBTEST((NAME<float, 4, ColMajor>())); \
-  CALL_SUBTEST((NAME<float, 5, ColMajor>()))
+  CALL_SUBTEST((NAME<float, 5, ColMajor>())); \
+  CALL_SUBTEST((NAME<bool, 1, RowMajor>())); \
+  CALL_SUBTEST((NAME<bool, 2, RowMajor>())); \
+  CALL_SUBTEST((NAME<bool, 4, RowMajor>())); \
+  CALL_SUBTEST((NAME<bool, 5, RowMajor>())); \
+  CALL_SUBTEST((NAME<bool, 1, ColMajor>())); \
+  CALL_SUBTEST((NAME<bool, 2, ColMajor>())); \
+  CALL_SUBTEST((NAME<bool, 4, ColMajor>())); \
+  CALL_SUBTEST((NAME<bool, 5, ColMajor>()))
 
 EIGEN_DECLARE_TEST(cxx11_tensor_block_io) {
   // clang-format off
diff --git a/unsupported/test/cxx11_tensor_contraction.cpp b/unsupported/test/cxx11_tensor_contraction.cpp
index 2fd128121..3b5c6a13c 100644
--- a/unsupported/test/cxx11_tensor_contraction.cpp
+++ b/unsupported/test/cxx11_tensor_contraction.cpp
@@ -562,36 +562,40 @@ static void test_large_contraction_with_output_kernel() {
 
 EIGEN_DECLARE_TEST(cxx11_tensor_contraction)
 {
-  CALL_SUBTEST(test_evals<ColMajor>());
-  CALL_SUBTEST(test_evals<RowMajor>());
-  CALL_SUBTEST(test_scalar<ColMajor>());
-  CALL_SUBTEST(test_scalar<RowMajor>());
-  CALL_SUBTEST(test_multidims<ColMajor>());
-  CALL_SUBTEST(test_multidims<RowMajor>());
-  CALL_SUBTEST(test_holes<ColMajor>());
-  CALL_SUBTEST(test_holes<RowMajor>());
-  CALL_SUBTEST(test_full_redux<ColMajor>());
-  CALL_SUBTEST(test_full_redux<RowMajor>());
-  CALL_SUBTEST(test_contraction_of_contraction<ColMajor>());
-  CALL_SUBTEST(test_contraction_of_contraction<RowMajor>());
-  CALL_SUBTEST(test_expr<ColMajor>());
-  CALL_SUBTEST(test_expr<RowMajor>());
-  CALL_SUBTEST(test_out_of_order_contraction<ColMajor>());
-  CALL_SUBTEST(test_out_of_order_contraction<RowMajor>());
-  CALL_SUBTEST(test_consistency<ColMajor>());
-  CALL_SUBTEST(test_consistency<RowMajor>());
-  CALL_SUBTEST(test_large_contraction<ColMajor>());
-  CALL_SUBTEST(test_large_contraction<RowMajor>());
-  CALL_SUBTEST(test_matrix_vector<ColMajor>());
-  CALL_SUBTEST(test_matrix_vector<RowMajor>());
-  CALL_SUBTEST(test_tensor_vector<ColMajor>());
-  CALL_SUBTEST(test_tensor_vector<RowMajor>());
-  CALL_SUBTEST(test_small_blocking_factors<ColMajor>());
-  CALL_SUBTEST(test_small_blocking_factors<RowMajor>());
-  CALL_SUBTEST(test_tensor_product<ColMajor>());
-  CALL_SUBTEST(test_tensor_product<RowMajor>());
-  CALL_SUBTEST(test_const_inputs<ColMajor>());
-  CALL_SUBTEST(test_const_inputs<RowMajor>());
-  CALL_SUBTEST(test_large_contraction_with_output_kernel<ColMajor>());
-  CALL_SUBTEST(test_large_contraction_with_output_kernel<RowMajor>());
+  CALL_SUBTEST_1(test_evals<ColMajor>());
+  CALL_SUBTEST_1(test_evals<RowMajor>());
+  CALL_SUBTEST_1(test_scalar<ColMajor>());
+  CALL_SUBTEST_1(test_scalar<RowMajor>());
+  CALL_SUBTEST_2(test_multidims<ColMajor>());
+  CALL_SUBTEST_2(test_multidims<RowMajor>());
+  CALL_SUBTEST_2(test_holes<ColMajor>());
+  CALL_SUBTEST_2(test_holes<RowMajor>());
+  CALL_SUBTEST_3(test_full_redux<ColMajor>());
+  CALL_SUBTEST_3(test_full_redux<RowMajor>());
+  CALL_SUBTEST_3(test_contraction_of_contraction<ColMajor>());
+  CALL_SUBTEST_3(test_contraction_of_contraction<RowMajor>());
+  CALL_SUBTEST_4(test_expr<ColMajor>());
+  CALL_SUBTEST_4(test_expr<RowMajor>());
+  CALL_SUBTEST_4(test_out_of_order_contraction<ColMajor>());
+  CALL_SUBTEST_4(test_out_of_order_contraction<RowMajor>());
+  CALL_SUBTEST_5(test_consistency<ColMajor>());
+  CALL_SUBTEST_5(test_consistency<RowMajor>());
+  CALL_SUBTEST_5(test_large_contraction<ColMajor>());
+  CALL_SUBTEST_5(test_large_contraction<RowMajor>());
+  CALL_SUBTEST_6(test_matrix_vector<ColMajor>());
+  CALL_SUBTEST_6(test_matrix_vector<RowMajor>());
+  CALL_SUBTEST_6(test_tensor_vector<ColMajor>());
+  CALL_SUBTEST_6(test_tensor_vector<RowMajor>());
+  CALL_SUBTEST_7(test_small_blocking_factors<ColMajor>());
+  CALL_SUBTEST_7(test_small_blocking_factors<RowMajor>());
+  CALL_SUBTEST_7(test_tensor_product<ColMajor>());
+  CALL_SUBTEST_7(test_tensor_product<RowMajor>());
+  CALL_SUBTEST_8(test_const_inputs<ColMajor>());
+  CALL_SUBTEST_8(test_const_inputs<RowMajor>());
+  CALL_SUBTEST_8(test_large_contraction_with_output_kernel<ColMajor>());
+  CALL_SUBTEST_8(test_large_contraction_with_output_kernel<RowMajor>());
+
+  // Force CMake to split this test.
+  // EIGEN_SUFFIXES;1;2;3;4;5;6;7;8
+
 }
diff --git a/unsupported/test/cxx11_tensor_expr.cpp b/unsupported/test/cxx11_tensor_expr.cpp
index d56da28d8..6dc44996a 100644
--- a/unsupported/test/cxx11_tensor_expr.cpp
+++ b/unsupported/test/cxx11_tensor_expr.cpp
@@ -195,26 +195,23 @@ static void test_constants()
 
 static void test_boolean()
 {
-  Tensor<int, 1> vec(31);
-  std::iota(vec.data(), vec.data() + 31, 0);
+  const int kSize = 31;
+  Tensor<int, 1> vec(kSize);
+  std::iota(vec.data(), vec.data() + kSize, 0);
 
   // Test ||.
   Tensor<bool, 1> bool1 = vec < vec.constant(1) || vec > vec.constant(4);
-  VERIFY_IS_EQUAL(bool1[0], true);
-  VERIFY_IS_EQUAL(bool1[1], false);
-  VERIFY_IS_EQUAL(bool1[2], false);
-  VERIFY_IS_EQUAL(bool1[3], false);
-  VERIFY_IS_EQUAL(bool1[4], false);
-  VERIFY_IS_EQUAL(bool1[5], true);
+  for (int i = 0; i < kSize; ++i) {
+    bool expected = i < 1 || i > 4;
+    VERIFY_IS_EQUAL(bool1[i], expected);
+  }
 
   // Test &&, including cast of operand vec.
   Tensor<bool, 1> bool2 = vec.cast<bool>() && vec < vec.constant(4);
-  VERIFY_IS_EQUAL(bool2[0], false);
-  VERIFY_IS_EQUAL(bool2[1], true);
-  VERIFY_IS_EQUAL(bool2[2], true);
-  VERIFY_IS_EQUAL(bool2[3], true);
-  VERIFY_IS_EQUAL(bool2[4], false);
-  VERIFY_IS_EQUAL(bool2[5], false);
+  for (int i = 0; i < kSize; ++i) {
+    bool expected = bool(i) && i < 4;
+    VERIFY_IS_EQUAL(bool2[i], expected);
+  }
 
   // Compilation tests:
   // Test Tensor<bool> against results of cast or comparison; verifies that
diff --git a/unsupported/test/cxx11_tensor_morphing.cpp b/unsupported/test/cxx11_tensor_morphing.cpp
index eb708737d..f01b95357 100644
--- a/unsupported/test/cxx11_tensor_morphing.cpp
+++ b/unsupported/test/cxx11_tensor_morphing.cpp
@@ -64,7 +64,7 @@ static void test_static_reshape() {
 #endif
 }
 
-template<typename>
+template <typename>
 static void test_reshape_in_expr() {
   MatrixXf m1(2,3*5*7*11);
   MatrixXf m2(3*5*7*11,13);
@@ -113,19 +113,19 @@ static void test_reshape_as_lvalue()
   }
 }
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_simple_slice()
 {
-  Tensor<float, 5, DataLayout> tensor(2,3,5,7,11);
+  Tensor<T, 5, DataLayout> tensor(2,3,5,7,11);
   tensor.setRandom();
 
-  Tensor<float, 5, DataLayout> slice1(1,1,1,1,1);
+  Tensor<T, 5, DataLayout> slice1(1,1,1,1,1);
   Eigen::DSizes<ptrdiff_t, 5> indices(1,2,3,4,5);
   Eigen::DSizes<ptrdiff_t, 5> sizes(1,1,1,1,1);
   slice1 = tensor.slice(indices, sizes);
   VERIFY_IS_EQUAL(slice1(0,0,0,0,0), tensor(1,2,3,4,5));
 
-  Tensor<float, 5, DataLayout> slice2(1,1,2,2,3);
+  Tensor<T, 5, DataLayout> slice2(1,1,2,2,3);
   Eigen::DSizes<ptrdiff_t, 5> indices2(1,1,3,4,5);
   Eigen::DSizes<ptrdiff_t, 5> sizes2(1,1,2,2,3);
   slice2 = tensor.slice(indices2, sizes2);
@@ -138,20 +138,20 @@ static void test_simple_slice()
   }
 }
 
-template<typename=void>
+template<typename T>
 static void test_const_slice()
 {
-  const float b[1] = {42};
-  TensorMap<Tensor<const float, 1> > m(b, 1);
+  const T b[1] = {42};
+  TensorMap<Tensor<const T, 1> > m(b, 1);
   DSizes<DenseIndex, 1> offsets;
   offsets[0] = 0;
-  TensorRef<Tensor<const float, 1> > slice_ref(m.slice(offsets, m.dimensions()));
+  TensorRef<Tensor<const T, 1> > slice_ref(m.slice(offsets, m.dimensions()));
   VERIFY_IS_EQUAL(slice_ref(0), 42);
 }
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_slice_in_expr() {
-  typedef Matrix<float, Dynamic, Dynamic, DataLayout> Mtx;
+  typedef Matrix<T, Dynamic, Dynamic, DataLayout> Mtx;
   Mtx m1(7,7);
   Mtx m2(3,3);
   m1.setRandom();
@@ -159,10 +159,10 @@ static void test_slice_in_expr() {
 
   Mtx m3 = m1.block(1, 2, 3, 3) * m2.block(0, 2, 3, 1);
 
-  TensorMap<Tensor<float, 2, DataLayout>> tensor1(m1.data(), 7, 7);
-  TensorMap<Tensor<float, 2, DataLayout>> tensor2(m2.data(), 3, 3);
-  Tensor<float, 2, DataLayout> tensor3(3,1);
-  typedef Tensor<float, 1>::DimensionPair DimPair;
+  TensorMap<Tensor<T, 2, DataLayout>> tensor1(m1.data(), 7, 7);
+  TensorMap<Tensor<T, 2, DataLayout>> tensor2(m2.data(), 3, 3);
+  Tensor<T, 2, DataLayout> tensor3(3,1);
+  typedef typename Tensor<T, 1>::DimensionPair DimPair;
   array<DimPair, 1> contract_along{{DimPair(1, 0)}};
 
   Eigen::DSizes<ptrdiff_t, 2> indices1(1,2);
@@ -179,28 +179,28 @@ static void test_slice_in_expr() {
   }
 
   // Take an arbitrary slice of an arbitrarily sized tensor.
-  TensorMap<Tensor<const float, 2, DataLayout>> tensor4(m1.data(), 7, 7);
-  Tensor<float, 1, DataLayout> tensor6 = tensor4.reshape(DSizes<ptrdiff_t, 1>(7*7)).exp().slice(DSizes<ptrdiff_t, 1>(0), DSizes<ptrdiff_t, 1>(35));
+  TensorMap<Tensor<const T, 2, DataLayout>> tensor4(m1.data(), 7, 7);
+  Tensor<T, 1, DataLayout> tensor6 = tensor4.reshape(DSizes<ptrdiff_t, 1>(7*7)).exp().slice(DSizes<ptrdiff_t, 1>(0), DSizes<ptrdiff_t, 1>(35));
   for (int i = 0; i < 35; ++i) {
     VERIFY_IS_APPROX(tensor6(i), expf(tensor4.data()[i]));
   }
 }
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_slice_as_lvalue()
 {
-  Tensor<float, 3, DataLayout> tensor1(2,2,7);
+  Tensor<T, 3, DataLayout> tensor1(2,2,7);
   tensor1.setRandom();
-  Tensor<float, 3, DataLayout> tensor2(2,2,7);
+  Tensor<T, 3, DataLayout> tensor2(2,2,7);
   tensor2.setRandom();
-  Tensor<float, 3, DataLayout> tensor3(4,3,5);
+  Tensor<T, 3, DataLayout> tensor3(4,3,5);
   tensor3.setRandom();
-  Tensor<float, 3, DataLayout> tensor4(4,3,2);
+  Tensor<T, 3, DataLayout> tensor4(4,3,2);
   tensor4.setRandom();
-  Tensor<float, 3, DataLayout> tensor5(10,13,12);
+  Tensor<T, 3, DataLayout> tensor5(10,13,12);
   tensor5.setRandom();
 
-  Tensor<float, 3, DataLayout> result(4,5,7);
+  Tensor<T, 3, DataLayout> result(4,5,7);
   Eigen::DSizes<ptrdiff_t, 3> sizes12(2,2,7);
   Eigen::DSizes<ptrdiff_t, 3> first_slice(0,0,0);
   result.slice(first_slice, sizes12) = tensor1;
@@ -246,10 +246,10 @@ static void test_slice_as_lvalue()
   }
 }
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_slice_raw_data()
 {
-  Tensor<float, 4, DataLayout> tensor(3,5,7,11);
+  Tensor<T, 4, DataLayout> tensor(3,5,7,11);
   tensor.setRandom();
 
   Eigen::DSizes<ptrdiff_t, 4> offsets(1,2,3,4);
@@ -276,7 +276,7 @@ static void test_slice_raw_data()
   extents = Eigen::DSizes<ptrdiff_t, 4>(1,2,1,1);
   auto slice3 = SliceEvaluator(tensor.slice(offsets, extents), DefaultDevice());
   VERIFY_IS_EQUAL(slice3.dimensions().TotalSize(), 2);
-  VERIFY_IS_EQUAL(slice3.data(), static_cast<float*>(0));
+  VERIFY_IS_EQUAL(slice3.data(), static_cast<T*>(0));
 
   if (DataLayout == ColMajor) {
     offsets = Eigen::DSizes<ptrdiff_t, 4>(0,2,3,4);
@@ -341,15 +341,15 @@ static void test_slice_raw_data()
 }
 
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_strided_slice()
 {
-  typedef Tensor<float, 5, DataLayout> Tensor5f;
+  typedef Tensor<T, 5, DataLayout> Tensor5f;
   typedef Eigen::DSizes<Eigen::DenseIndex, 5> Index5;
-  typedef Tensor<float, 2, DataLayout> Tensor2f;
+  typedef Tensor<T, 2, DataLayout> Tensor2f;
   typedef Eigen::DSizes<Eigen::DenseIndex, 2> Index2;
-  Tensor<float, 5, DataLayout> tensor(2,3,5,7,11);
-  Tensor<float, 2, DataLayout> tensor2(7,11);
+  Tensor<T, 5, DataLayout> tensor(2,3,5,7,11);
+  Tensor<T, 2, DataLayout> tensor2(7,11);
   tensor.setRandom();
   tensor2.setRandom();
 
@@ -435,13 +435,13 @@ static void test_strided_slice()
   }
 }
 
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_strided_slice_write()
 {
-  typedef Tensor<float, 2, DataLayout> Tensor2f;
+  typedef Tensor<T, 2, DataLayout> Tensor2f;
   typedef Eigen::DSizes<Eigen::DenseIndex, 2> Index2;
 
-  Tensor<float, 2, DataLayout> tensor(7,11),tensor2(7,11);
+  Tensor<T, 2, DataLayout> tensor(7,11),tensor2(7,11);
   tensor.setRandom();
   tensor2=tensor;
   Tensor2f slice(2,3);
@@ -461,15 +461,14 @@ static void test_strided_slice_write()
   }
 }
 
-
-template<int DataLayout>
+template<typename T, int DataLayout>
 static void test_composition()
 {
-  Eigen::Tensor<float, 2, DataLayout> matrix(7, 11);
+  Eigen::Tensor<T, 2, DataLayout> matrix(7, 11);
   matrix.setRandom();
 
   const DSizes<ptrdiff_t, 3> newDims(1, 1, 11);
-  Eigen::Tensor<float, 3, DataLayout> tensor =
+  Eigen::Tensor<T, 3, DataLayout> tensor =
       matrix.slice(DSizes<ptrdiff_t, 2>(2, 0), DSizes<ptrdiff_t, 2>(1, 11)).reshape(newDims);
 
   VERIFY_IS_EQUAL(tensor.dimensions().TotalSize(), 11);
@@ -481,29 +480,27 @@ static void test_composition()
   }
 }
 
+#define CALL_SUBTEST_PART(PART) \
+  CALL_SUBTEST_##PART
+
+#define CALL_SUBTESTS_TYPES_LAYOUTS(PART, NAME)       \
+  CALL_SUBTEST_PART(PART)((NAME<float, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<float, RowMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, ColMajor>())); \
+  CALL_SUBTEST_PART(PART)((NAME<bool, RowMajor>()))
 
 EIGEN_DECLARE_TEST(cxx11_tensor_morphing)
 {
   CALL_SUBTEST_1(test_simple_reshape<void>());
   CALL_SUBTEST_1(test_static_reshape<void>());
-  CALL_SUBTEST_1(test_reshape_in_expr<void>());
   CALL_SUBTEST_1(test_reshape_as_lvalue<void>());
-
-  CALL_SUBTEST_1(test_simple_slice<ColMajor>());
-  CALL_SUBTEST_1(test_simple_slice<RowMajor>());
-  CALL_SUBTEST_1(test_const_slice());
-  CALL_SUBTEST_2(test_slice_in_expr<ColMajor>());
-  CALL_SUBTEST_3(test_slice_in_expr<RowMajor>());
-  CALL_SUBTEST_4(test_slice_as_lvalue<ColMajor>());
-  CALL_SUBTEST_4(test_slice_as_lvalue<RowMajor>());
-  CALL_SUBTEST_5(test_slice_raw_data<ColMajor>());
-  CALL_SUBTEST_5(test_slice_raw_data<RowMajor>());
-
-  CALL_SUBTEST_6(test_strided_slice_write<ColMajor>());
-  CALL_SUBTEST_6(test_strided_slice<ColMajor>());
-  CALL_SUBTEST_6(test_strided_slice_write<RowMajor>());
-  CALL_SUBTEST_6(test_strided_slice<RowMajor>());
-
-  CALL_SUBTEST_7(test_composition<ColMajor>());
-  CALL_SUBTEST_7(test_composition<RowMajor>());
+  CALL_SUBTEST_1(test_reshape_in_expr<void>());
+  CALL_SUBTEST_1(test_const_slice<float>());
+
+  CALL_SUBTESTS_TYPES_LAYOUTS(2, test_simple_slice);
+  CALL_SUBTESTS_TYPES_LAYOUTS(3, test_slice_as_lvalue);
+  CALL_SUBTESTS_TYPES_LAYOUTS(4, test_slice_raw_data);
+  CALL_SUBTESTS_TYPES_LAYOUTS(5, test_strided_slice_write);
+  CALL_SUBTESTS_TYPES_LAYOUTS(6, test_strided_slice);
+  CALL_SUBTESTS_TYPES_LAYOUTS(7, test_composition);
 }