COMPMID-1933: Implement NEHeightConcatenateLayer.

Added support to concactenate tensors along the Y axis in NEConcatenateLayer.

Change-Id: Ib714bfcf9954cc35918efa7d52fc9164bb08bdf6
Signed-off-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/841
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

10 files changed, 266 insertions, 77 deletions

diff --git a/tests/datasets/ShapeDatasets.h b/tests/datasets/ShapeDatasets.h
index fddc7731fe..b6923c15c9 100644
--- a/tests/datasets/ShapeDatasets.h
+++ b/tests/datasets/ShapeDatasets.h
@@ -732,11 +732,11 @@ public:
     }
 };
 
-/** Data set containing tensor shapes for WidthConcatenateLayer. */
-class WidthConcatenateLayerShapes final : public ShapeDataset
+/** Data set containing tensor shapes for ConcatenateLayer. */
+class ConcatenateLayerShapes final : public ShapeDataset
 {
 public:
-    WidthConcatenateLayerShapes()
+    ConcatenateLayerShapes()
         : ShapeDataset("Shape",
     {
         TensorShape{ 232U, 65U, 3U },
diff --git a/tests/validation/CL/WidthConcatenateLayer.cpp b/tests/validation/CL/WidthConcatenateLayer.cpp
index 647e0413a1..493320b9ad 100644
--- a/tests/validation/CL/WidthConcatenateLayer.cpp
+++ b/tests/validation/CL/WidthConcatenateLayer.cpp
@@ -24,14 +24,14 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "arm_compute/runtime/CL/functions/CLWidthConcatenateLayer.h"
+#include "arm_compute/runtime/CL/functions/CLConcatenateLayer.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/datasets/ShapeDatasets.h"
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/WidthConcatenateLayerFixture.h"
+#include "tests/validation/fixtures/ConcatenateLayerFixture.h"
 
 namespace arm_compute
 {
@@ -72,8 +72,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
         inputs_vector_info_raw.emplace_back(&input);
     }
 
-    bool is_valid = bool(CLWidthConcatenateLayer::validate(inputs_vector_info_raw,
-                                                           &output_info.clone()->set_is_resizable(false)));
+    bool is_valid = bool(CLConcatenateLayer::validate(inputs_vector_info_raw,&output_info.clone()->set_is_resizable(false),DataLayoutDimension::WIDTH ));
     ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS);
 }
 // clang-format on
@@ -93,26 +92,30 @@ TEST_CASE(Configuration, framework::DatasetMode::ALL)
     ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
 
     // Create and configure function
-    CLWidthConcatenateLayer concat_layer;
+    CLConcatenateLayer concat_layer;
 
-    concat_layer.configure({ &src1, &src2, &src3 }, &dst);
+    concat_layer.configure({ &src1, &src2, &src3 }, &dst, DataLayoutDimension::WIDTH);
 }
 
 template <typename T>
-using CLWidthConcatenateLayerFixture = WidthConcatenateLayerValidationFixture<CLTensor, ICLTensor, CLAccessor, CLWidthConcatenateLayer, T>;
+using CLWidthConcatenateLayerFixture = ConcatenateLayerValidationFixture<CLTensor, ICLTensor, CLAccessor, CLConcatenateLayer, T>;
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
                                                                                                                   framework::dataset::make("DataType",
-                                                                                                                          DataType::F16)))
+                                                                                                                          DataType::F16)),
+                                                                                                                  framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<half>, framework::DatasetMode::NIGHTLY, combine(concat(datasets::Large2DShapes(), datasets::Small4DShapes()),
-                                                                                                                framework::dataset::make("DataType",
-                                                                                                                        DataType::F16)))
+FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<half>, framework::DatasetMode::NIGHTLY, combine(combine(concat(datasets::Large2DShapes(), datasets::Small4DShapes()),
+                                                                                                                        framework::dataset::make("DataType",
+                                                                                                                                DataType::F16)),
+                                                                                                                framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -120,15 +123,18 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<half>, framework
 TEST_SUITE_END()
 
 TEST_SUITE(FP32)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
                                                                                                                    framework::dataset::make("DataType",
-                                                                                                                           DataType::F32)))
+                                                                                                                           DataType::F32)),
+                                                                                                                   framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::WidthConcatenateLayerShapes(), framework::dataset::make("DataType",
-                                                                                                                 DataType::F32)))
+FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(), framework::dataset::make("DataType",
+                                                                                                                 DataType::F32)),
+                                                                                                                 framework::dataset::make("Axis", 0)))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -138,15 +144,17 @@ TEST_SUITE_END()
 
 TEST_SUITE(Quantized)
 TEST_SUITE(QASYMM8)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
                                                                                                                      framework::dataset::make("DataType",
-                                                                                                                             DataType::QASYMM8)))
+                                                                                                                             DataType::QASYMM8)),
+                                                                                                                     framework::dataset::make("Axis", 0)))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(datasets::WidthConcatenateLayerShapes(), framework::dataset::make("DataType",
-                                                                                                                   DataType::QASYMM8)))
+FIXTURE_DATA_TEST_CASE(RunLarge, CLWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(), framework::dataset::make("DataType",
+                                                                                                                   DataType::QASYMM8)),
+                                                                                                                   framework::dataset::make("Axis", 0)))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
diff --git a/tests/validation/Helpers.cpp b/tests/validation/Helpers.cpp
index 11c454ea67..e9612f2223 100644
--- a/tests/validation/Helpers.cpp
+++ b/tests/validation/Helpers.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -98,17 +98,18 @@ TensorShape calculate_depth_concatenate_shape(const std::vector<TensorShape> &in
     return out_shape;
 }
 
-TensorShape calculate_width_concatenate_shape(const std::vector<TensorShape> &input_shapes)
+TensorShape calculate_concatenate_shape(const std::vector<TensorShape> &input_shapes, size_t axis)
 {
     ARM_COMPUTE_ERROR_ON(input_shapes.empty());
-
     TensorShape out_shape = input_shapes[0];
+    ARM_COMPUTE_ERROR_ON(axis >= out_shape.num_dimensions());
 
-    int width = std::accumulate(input_shapes.begin(), input_shapes.end(), 0, [](int sum, const TensorShape & shape)
+    const int new_size = std::accumulate(input_shapes.begin(), input_shapes.end(), 0, [&](int sum, const TensorShape & shape)
     {
-        return sum + shape.x();
+        ARM_COMPUTE_ERROR_ON(axis >= shape.num_dimensions());
+        return sum + shape[axis];
     });
-    out_shape.set(0, width);
+    out_shape.set(axis, new_size);
 
     return out_shape;
 }
diff --git a/tests/validation/Helpers.h b/tests/validation/Helpers.h
index 4d1d21440d..2e8c667a41 100644
--- a/tests/validation/Helpers.h
+++ b/tests/validation/Helpers.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -128,13 +128,14 @@ void fill_mask_from_pattern(uint8_t *mask, int cols, int rows, MatrixPattern pat
  */
 TensorShape calculate_depth_concatenate_shape(const std::vector<TensorShape> &input_shapes);
 
-/** Calculate output tensor shape give a vector of input tensor to concatenate
+/** Calculate output tensor shape for the concatenate operation along a given axis
  *
  * @param[in] input_shapes Shapes of the tensors to concatenate across width.
+ * @param[in] axis         Axis to use for the concatenate operation
  *
  * @return The shape of output concatenated tensor.
  */
-TensorShape calculate_width_concatenate_shape(const std::vector<TensorShape> &input_shapes);
+TensorShape calculate_concatenate_shape(const std::vector<TensorShape> &input_shapes, size_t axis);
 
 /** Parameters of Harris Corners algorithm. */
 struct HarrisCornersParameters
diff --git a/tests/validation/NEON/HeightConcatenateLayer.cpp b/tests/validation/NEON/HeightConcatenateLayer.cpp
new file mode 100644
index 0000000000..f5400f9246
--- /dev/null
+++ b/tests/validation/NEON/HeightConcatenateLayer.cpp
@@ -0,0 +1,131 @@
+/*
+ * Copyright (c) 2019 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/functions/NEConcatenateLayer.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "arm_compute/runtime/TensorAllocator.h"
+#include "tests/NEON/Accessor.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ConcatenateLayerFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+TEST_SUITE(NEON)
+TEST_SUITE(HeightConcatenateLayer)
+
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
+        framework::dataset::make("InputInfo1", {  TensorInfo(TensorShape(23U, 15U, 5U), 1, DataType::F32), // Mismatching data type input/output
+                                                  TensorInfo(TensorShape(22U, 27U, 5U), 1, DataType::F32), // Mismatching y dimension
+                                                  TensorInfo(TensorShape(11U, 25U, 5U), 1, DataType::F32), // Mismatching total height
+                                                  TensorInfo(TensorShape(16U, 25U, 5U), 1, DataType::F32)
+        }),
+        framework::dataset::make("InputInfo2", {  TensorInfo(TensorShape(23U, 15U, 4U), 1, DataType::F32),
+                                                  TensorInfo(TensorShape(22U, 127U, 5U), 1, DataType::F32),
+                                                  TensorInfo(TensorShape(11U, 26U, 5U), 1, DataType::F32),
+                                                  TensorInfo(TensorShape(16U, 25U, 5U), 1, DataType::F32)
+        })),
+        framework::dataset::make("OutputInfo", {  TensorInfo(TensorShape(23U, 30U, 5U), 1, DataType::F16),
+                                                  TensorInfo(TensorShape(22U, 12U, 5U), 1, DataType::F32),
+                                                  TensorInfo(TensorShape(11U, 7U, 5U), 1, DataType::F32),
+                                                  TensorInfo(TensorShape(16U, 50U, 5U), 1, DataType::F32)
+        })),
+        framework::dataset::make("Expected", { false, false, false, true })),
+        input_info1, input_info2, output_info,expected)
+{
+    std::vector<TensorInfo> inputs_vector_info;
+    inputs_vector_info.emplace_back(std::move(input_info1));
+    inputs_vector_info.emplace_back(std::move(input_info2));
+
+    std::vector<ITensorInfo *> inputs_vector_info_raw;
+    for(auto &input : inputs_vector_info)
+    {
+        inputs_vector_info_raw.emplace_back(&input);
+    }
+
+    bool is_valid = bool(NEConcatenateLayer::validate(inputs_vector_info_raw, &output_info.clone()->set_is_resizable(false), DataLayoutDimension::HEIGHT));
+    ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using NEHeightConcatenateLayerFixture = ConcatenateLayerValidationFixture<Tensor, ITensor, Accessor, NEConcatenateLayer, T>;
+
+TEST_SUITE(Float)
+TEST_SUITE(FP32)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEHeightConcatenateLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+                                                                                                                    framework::dataset::make("DataType",
+                                                                                                                            DataType::F32)),
+                                                                                                                    framework::dataset::make("Axis", 1)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEHeightConcatenateLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(), framework::dataset::make("DataType",
+                                                                                                                  DataType::F32)),
+                                                                                                                  framework::dataset::make("Axis", 1)))
+
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // FP32
+TEST_SUITE_END() // Float
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEHeightConcatenateLayerFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+                                                                                                                      framework::dataset::make("DataType",
+                                                                                                                              DataType::QASYMM8)),
+                                                                                                                      framework::dataset::make("Axis", 1)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEHeightConcatenateLayerFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(),
+                                                                                                                    framework::dataset::make("DataType",
+                                                                                                                            DataType::QASYMM8)),
+                                                                                                                    framework::dataset::make("Axis", 1)))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // QASYMM8
+TEST_SUITE_END() // Quantized
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/NEON/WidthConcatenateLayer.cpp b/tests/validation/NEON/WidthConcatenateLayer.cpp
index 6e94e92d05..dba14ebb35 100644
--- a/tests/validation/NEON/WidthConcatenateLayer.cpp
+++ b/tests/validation/NEON/WidthConcatenateLayer.cpp
@@ -22,7 +22,7 @@
  * SOFTWARE.
  */
 #include "arm_compute/core/Types.h"
-#include "arm_compute/runtime/NEON/functions/NEWidthConcatenateLayer.h"
+#include "arm_compute/runtime/NEON/functions/NEConcatenateLayer.h"
 #include "arm_compute/runtime/Tensor.h"
 #include "arm_compute/runtime/TensorAllocator.h"
 #include "tests/NEON/Accessor.h"
@@ -31,7 +31,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/WidthConcatenateLayerFixture.h"
+#include "tests/validation/fixtures/ConcatenateLayerFixture.h"
 
 namespace arm_compute
 {
@@ -72,27 +72,30 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
         inputs_vector_info_raw.emplace_back(&input);
     }
 
-    bool is_valid = bool(NEWidthConcatenateLayer::validate(inputs_vector_info_raw,
-                                                           &output_info.clone()->set_is_resizable(false)));
+    bool is_valid = bool(NEConcatenateLayer::validate(inputs_vector_info_raw,
+                                                           &output_info.clone()->set_is_resizable(false),DataLayoutDimension::WIDTH));
     ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS);
 }
 // clang-format on
 // *INDENT-ON*
-
 template <typename T>
-using NEWidthConcatenateLayerFixture = WidthConcatenateLayerValidationFixture<Tensor, ITensor, Accessor, NEWidthConcatenateLayer, T>;
+using NEWidthConcatenateLayerFixture = ConcatenateLayerValidationFixture<Tensor, ITensor, Accessor, NEConcatenateLayer, T>;
 
 TEST_SUITE(Float)
 TEST_SUITE(FP32)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEWidthConcatenateLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+FIXTURE_DATA_TEST_CASE(RunSmall, NEWidthConcatenateLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
                                                                                                                    framework::dataset::make("DataType",
-                                                                                                                           DataType::F32)))
+                                                                                                                           DataType::F32)),
+                                                                                                                   framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
-FIXTURE_DATA_TEST_CASE(RunLarge, NEWidthConcatenateLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::WidthConcatenateLayerShapes(), framework::dataset::make("DataType",
-                                                                                                                 DataType::F32)))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEWidthConcatenateLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(), framework::dataset::make("DataType",
+                                                                                                                 DataType::F32)),
+                                                                                                                 framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -102,15 +105,19 @@ TEST_SUITE_END()
 
 TEST_SUITE(Quantized)
 TEST_SUITE(QASYMM8)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
+FIXTURE_DATA_TEST_CASE(RunSmall, NEWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(concat(datasets::Small2DShapes(), datasets::Tiny4DShapes()),
                                                                                                                      framework::dataset::make("DataType",
-                                                                                                                             DataType::QASYMM8)))
+                                                                                                                             DataType::QASYMM8)),
+                                                                                                                     framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
-FIXTURE_DATA_TEST_CASE(RunLarge, NEWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(datasets::WidthConcatenateLayerShapes(), framework::dataset::make("DataType",
-                                                                                                                   DataType::QASYMM8)))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEWidthConcatenateLayerFixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::ConcatenateLayerShapes(), framework::dataset::make("DataType",
+                                                                                                                   DataType::QASYMM8)),
+                                                                                                                   framework::dataset::make("Axis", 0)))
+
 {
     // Validate output
     validate(Accessor(_target), _reference);
diff --git a/tests/validation/fixtures/WidthConcatenateLayerFixture.h b/tests/validation/fixtures/ConcatenateLayerFixture.h
index 47a03ed865..db09957c09 100644
--- a/tests/validation/fixtures/WidthConcatenateLayerFixture.h
+++ b/tests/validation/fixtures/ConcatenateLayerFixture.h
@@ -33,7 +33,7 @@
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Fixture.h"
 #include "tests/validation/Helpers.h"
-#include "tests/validation/reference/WidthConcatenateLayer.h"
+#include "tests/validation/reference/ConcatenateLayer.h"
 
 #include <random>
 
@@ -44,11 +44,11 @@ namespace test
 namespace validation
 {
 template <typename TensorType, typename ITensorType, typename AccessorType, typename FunctionType, typename T>
-class WidthConcatenateLayerValidationFixture : public framework::Fixture
+class ConcatenateLayerValidationFixture : public framework::Fixture
 {
 public:
     template <typename...>
-    void setup(TensorShape shape, DataType data_type)
+    void setup(TensorShape shape, DataType data_type, unsigned int axis)
     {
         // Create input shapes
         std::mt19937                    gen(library->seed());
@@ -78,12 +78,12 @@ public:
             {
                 // Decrease the dimension by a small percentage. Don't increase
                 // as that could make tensor too large.
-                s.set(0, s[0] + 2 * static_cast<int>(s[0] * change_dis(gen)));
+                s.set(axis, s[axis] + 2 * static_cast<int>(s[axis] * change_dis(gen)));
             }
         }
 
-        _target    = compute_target(shapes, qinfo, data_type);
-        _reference = compute_reference(shapes, qinfo, data_type);
+        _target    = compute_target(shapes, qinfo, data_type, axis);
+        _reference = compute_reference(shapes, qinfo, data_type, axis);
     }
 
 protected:
@@ -93,7 +93,7 @@ protected:
         library->fill_tensor_uniform(tensor, i);
     }
 
-    TensorType compute_target(std::vector<TensorShape> shapes, const std::vector<QuantizationInfo> &qinfo, DataType data_type)
+    TensorType compute_target(const std::vector<TensorShape> &shapes, const std::vector<QuantizationInfo> &qinfo, DataType data_type, unsigned int axis)
     {
         std::vector<TensorType>    srcs;
         std::vector<ITensorType *> src_ptrs;
@@ -107,13 +107,26 @@ protected:
             src_ptrs.emplace_back(&srcs.back());
         }
 
-        TensorShape dst_shape = misc::shape_calculator::calculate_width_concatenate_shape(src_ptrs);
-
-        TensorType dst = create_tensor<TensorType>(dst_shape, data_type, 1, qinfo[shapes.size()]);
+        const TensorShape dst_shape = misc::shape_calculator::calculate_concatenate_shape(src_ptrs, axis);
+        TensorType        dst       = create_tensor<TensorType>(dst_shape, data_type, 1, qinfo[shapes.size()]);
 
         // Create and configure function
-        FunctionType width_concat;
-        width_concat.configure(src_ptrs, &dst);
+        FunctionType concat;
+        switch(axis)
+        {
+            case 0:
+                concat.configure(src_ptrs, &dst, DataLayoutDimension::WIDTH);
+                break;
+            case 1:
+                concat.configure(src_ptrs, &dst, DataLayoutDimension::HEIGHT);
+                break;
+            case 2:
+                concat.configure(src_ptrs, &dst, DataLayoutDimension::CHANNEL);
+                break;
+            default:
+                ARM_COMPUTE_ERROR("Not supported");
+                break;
+        }
 
         for(auto &src : srcs)
         {
@@ -140,12 +153,12 @@ protected:
         }
 
         // Compute function
-        width_concat.run();
+        concat.run();
 
         return dst;
     }
 
-    SimpleTensor<T> compute_reference(std::vector<TensorShape> shapes, const std::vector<QuantizationInfo> &qinfo, DataType data_type)
+    SimpleTensor<T> compute_reference(const std::vector<TensorShape> &shapes, const std::vector<QuantizationInfo> &qinfo, DataType data_type, unsigned int axis)
     {
         std::vector<SimpleTensor<T>> srcs;
 
@@ -156,10 +169,9 @@ protected:
             fill(srcs.back(), j);
         }
 
-        const TensorShape dst_shape = calculate_width_concatenate_shape(shapes);
+        const TensorShape dst_shape = calculate_concatenate_shape(shapes, axis);
         SimpleTensor<T>   dst{ dst_shape, data_type, 1, qinfo[shapes.size()] };
-
-        return reference::widthconcatenate_layer<T>(srcs, dst);
+        return reference::concatenate_layer<T>(srcs, dst, axis);
     }
 
     TensorType      _target{};
diff --git a/tests/validation/fixtures/LSTMLayerFixture.h b/tests/validation/fixtures/LSTMLayerFixture.h
index b30f1e534b..2cf83b8b3d 100644
--- a/tests/validation/fixtures/LSTMLayerFixture.h
+++ b/tests/validation/fixtures/LSTMLayerFixture.h
@@ -29,11 +29,11 @@
 #include "tests/framework/Fixture.h"
 #include "tests/validation/reference/ActivationLayer.h"
 #include "tests/validation/reference/ArithmeticOperations.h"
+#include "tests/validation/reference/ConcatenateLayer.h"
 #include "tests/validation/reference/FullyConnectedLayer.h"
 #include "tests/validation/reference/GEMM.h"
 #include "tests/validation/reference/PixelWiseMultiplication.h"
 #include "tests/validation/reference/Transpose.h"
-#include "tests/validation/reference/WidthConcatenateLayer.h"
 
 namespace arm_compute
 {
@@ -415,7 +415,7 @@ protected:
         scratch_inputs.emplace_back(std::move(cell_state_out));
         scratch_inputs.emplace_back(std::move(forget_gate));
         scratch_inputs.emplace_back(std::move(output));
-        scratch            = reference::widthconcatenate_layer(scratch_inputs, scratch);
+        scratch            = reference::concatenate_layer(scratch_inputs, scratch, Window::DimX);
         _reference_scratch = std::move(scratch);
         return output_state_out;
     }
diff --git a/tests/validation/reference/WidthConcatenateLayer.cpp b/tests/validation/reference/ConcatenateLayer.cpp
index 38543393ce..1440878829 100644
--- a/tests/validation/reference/WidthConcatenateLayer.cpp
+++ b/tests/validation/reference/ConcatenateLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2019 ARM Limited.
+ * Copyright (c) 2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -21,9 +21,10 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "WidthConcatenateLayer.h"
+#include "ConcatenateLayer.h"
 
 #include "tests/validation/Helpers.h"
+#include "tests/validation/reference/Permute.h"
 
 namespace arm_compute
 {
@@ -33,24 +34,22 @@ namespace validation
 {
 namespace reference
 {
+namespace
+{
 template <typename T>
 SimpleTensor<T> widthconcatenate_layer(const std::vector<SimpleTensor<T>> &srcs, SimpleTensor<T> &dst)
 {
     // Create reference
     std::vector<TensorShape> shapes;
-
     for(const auto &src : srcs)
     {
         shapes.emplace_back(src.shape());
     }
-
     // Compute reference
     int       width_offset = 0;
     const int width_out    = dst.shape().x();
-
     // Set output tensor to 0
     std::fill_n(dst.data(), dst.num_elements(), 0);
-
     for(const auto &src : srcs)
     {
         ARM_COMPUTE_ERROR_ON(width_offset >= width_out);
@@ -89,13 +88,43 @@ SimpleTensor<T> widthconcatenate_layer(const std::vector<SimpleTensor<T>> &srcs,
         }
         width_offset += width;
     }
-
     return dst;
 }
 
 template SimpleTensor<float> widthconcatenate_layer(const std::vector<SimpleTensor<float>> &srcs, SimpleTensor<float> &dst);
 template SimpleTensor<half> widthconcatenate_layer(const std::vector<SimpleTensor<half>> &srcs, SimpleTensor<half> &dst);
 template SimpleTensor<uint8_t> widthconcatenate_layer(const std::vector<SimpleTensor<uint8_t>> &srcs, SimpleTensor<uint8_t> &dst);
+} // namespace
+
+template <typename T>
+SimpleTensor<T> concatenate_layer(std::vector<SimpleTensor<T>> &srcs, SimpleTensor<T> &dst, unsigned int axis)
+{
+    switch(axis)
+    {
+        case Window::DimX:
+        {
+            return widthconcatenate_layer(srcs, dst);
+        }
+        case Window::DimY:
+        {
+            for(auto &t : srcs)
+            {
+                t = reference::permute<T>(t, PermutationVector(1U, 0U));
+            }
+            dst = reference::permute<T>(dst, PermutationVector(1U, 0U));
+            return reference::permute<T>(widthconcatenate_layer(srcs, dst), PermutationVector(1U, 0U));
+        }
+        default:
+        {
+            ARM_COMPUTE_ERROR("Not supported");
+            return dst;
+        }
+    }
+}
+
+template SimpleTensor<float> concatenate_layer(std::vector<SimpleTensor<float>> &srcs, SimpleTensor<float> &dst, unsigned int axis);
+template SimpleTensor<half> concatenate_layer(std::vector<SimpleTensor<half>> &srcs, SimpleTensor<half> &dst, unsigned int axis);
+template SimpleTensor<uint8_t> concatenate_layer(std::vector<SimpleTensor<uint8_t>> &srcs, SimpleTensor<uint8_t> &dst, unsigned int axis);
 } // namespace reference
 } // namespace validation
 } // namespace test
diff --git a/tests/validation/reference/WidthConcatenateLayer.h b/tests/validation/reference/ConcatenateLayer.h
index 0f1f428f10..14fd097eee 100644
--- a/tests/validation/reference/WidthConcatenateLayer.h
+++ b/tests/validation/reference/ConcatenateLayer.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2019 ARM Limited.
+ * Copyright (c) 2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef __ARM_COMPUTE_TEST_WIDTHCONCATENATE_LAYER_H__
-#define __ARM_COMPUTE_TEST_WIDTHCONCATENATE_LAYER_H__
+#ifndef __ARM_COMPUTE_TEST_CONCATENATE_LAYER_H__
+#define __ARM_COMPUTE_TEST_CONCATENATE_LAYER_H__
 
 #include "tests/SimpleTensor.h"
 
@@ -37,9 +37,9 @@ namespace validation
 namespace reference
 {
 template <typename T>
-SimpleTensor<T> widthconcatenate_layer(const std::vector<SimpleTensor<T>> &srcs, SimpleTensor<T> &dst);
+SimpleTensor<T> concatenate_layer(std::vector<SimpleTensor<T>> &srcs, SimpleTensor<T> &dst, unsigned int axis);
 } // namespace reference
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
-#endif /* __ARM_COMPUTE_TEST_WIDTHCONCATENATE_LAYER_H__ */
+#endif /* __ARM_COMPUTE_TEST_CONCATENATE_LAYER_H__ */