Add CL/NEON DepthConcatenateLayer benchmark tests
Change-Id: Ibdfa0ad6e933a9b4b27504ead9bec999488c4636
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/113460
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
diff --git a/tests/benchmark/CL/DepthConcatenateLayer.cpp b/tests/benchmark/CL/DepthConcatenateLayer.cpp
new file mode 100644
index 0000000..ebc225a
--- /dev/null
+++ b/tests/benchmark/CL/DepthConcatenateLayer.cpp
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2017 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/TensorShape.h"
+#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/CLDepthConcatenateLayer.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/benchmark/fixtures/DepthConcatenateLayerFixture.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace benchmark
+{
+namespace
+{
+const auto data_types = framework::dataset::make("DataType", { DataType::QS8, DataType::QS16, DataType::F16, DataType::F32 });
+} // namespace
+
+using CLDepthConcatenateLayerFixture = DepthConcatenateLayerFixture<CLTensor, ICLTensor, CLDepthConcatenateLayer, CLAccessor>;
+
+TEST_SUITE(CL)
+TEST_SUITE(DepthConcatenateLayer)
+
+REGISTER_FIXTURE_DATA_TEST_CASE(RunSmall, CLDepthConcatenateLayerFixture, framework::DatasetMode::PRECOMMIT, framework::dataset::combine(datasets::Small2DShapes(), data_types));
+REGISTER_FIXTURE_DATA_TEST_CASE(RunLarge, CLDepthConcatenateLayerFixture, framework::DatasetMode::NIGHTLY, framework::dataset::combine(datasets::DepthConcatenateLayerShapes(), data_types));
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace benchmark
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/benchmark/NEON/DepthConcatenateLayer.cpp b/tests/benchmark/NEON/DepthConcatenateLayer.cpp
new file mode 100644
index 0000000..129dc60
--- /dev/null
+++ b/tests/benchmark/NEON/DepthConcatenateLayer.cpp
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2017 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/TensorShape.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/runtime/NEON/functions/NEDepthConcatenateLayer.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "arm_compute/runtime/TensorAllocator.h"
+#include "tests/NEON/Accessor.h"
+#include "tests/benchmark/fixtures/DepthConcatenateLayerFixture.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "utils/TypePrinter.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace benchmark
+{
+namespace
+{
+const auto data_types = framework::dataset::make("DataType", { DataType::QS8, DataType::QS16, DataType::F16, DataType::F32 });
+} // namespace
+
+using NEDepthConcatenateLayerFixture = DepthConcatenateLayerFixture<Tensor, ITensor, NEDepthConcatenateLayer, Accessor>;
+
+TEST_SUITE(NE)
+TEST_SUITE(DepthConcatenateLayer)
+
+REGISTER_FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConcatenateLayerFixture, framework::DatasetMode::PRECOMMIT, framework::dataset::combine(datasets::Small2DShapes(), data_types));
+REGISTER_FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConcatenateLayerFixture, framework::DatasetMode::NIGHTLY, framework::dataset::combine(datasets::DepthConcatenateLayerShapes(), data_types));
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace benchmark
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/benchmark/fixtures/DepthConcatenateLayerFixture.h b/tests/benchmark/fixtures/DepthConcatenateLayerFixture.h
new file mode 100644
index 0000000..221f34c
--- /dev/null
+++ b/tests/benchmark/fixtures/DepthConcatenateLayerFixture.h
@@ -0,0 +1,147 @@
+/*
+ * Copyright (c) 2017 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.
+ */
+#ifndef ARM_COMPUTE_TEST_DEPTHCONCATENATELAYERFIXTURE
+#define ARM_COMPUTE_TEST_DEPTHCONCATENATELAYERFIXTURE
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Utils.h"
+#include "tests/AssetsLibrary.h"
+#include "tests/Globals.h"
+#include "tests/Utils.h"
+#include "tests/framework/Fixture.h"
+
+#include <random>
+
+namespace arm_compute
+{
+namespace test
+{
+namespace benchmark
+{
+/** Fixture that can be used for NE/CL/GC */
+template <typename TensorType, typename ITensorType, typename Function, typename AccessorType>
+class DepthConcatenateLayerFixture : public framework::Fixture
+{
+public:
+ inline std::vector<TensorShape> generate_input_shapes(TensorShape shape)
+ {
+ // Create input shapes
+ std::mt19937 gen(library->seed());
+ std::uniform_int_distribution<> num_dis(2, 6);
+ const int num_tensors = num_dis(gen);
+
+ std::vector<TensorShape> shapes(num_tensors, shape);
+ std::uniform_int_distribution<> depth_dis(1, 7);
+ std::bernoulli_distribution mutate_dis(0.25f);
+ std::uniform_real_distribution<> change_dis(-0.25f, 0.f);
+
+ // Generate more shapes based on the input
+ for(auto &s : shapes)
+ {
+ // Set the depth of the tensor
+ s.set(2, depth_dis(gen));
+
+ // Randomly change the first dimension
+ if(mutate_dis(gen))
+ {
+ // Decrease the dimension by a small percentage. Don't increase
+ // as that could make tensor too large. Also the change must be
+ // an even number. Otherwise out depth concatenate fails.
+ s.set(0, s[0] + 2 * static_cast<int>(s[0] * change_dis(gen)));
+ }
+
+ // Repeat the same as above for the second dimension
+ if(mutate_dis(gen))
+ {
+ s.set(1, s[1] + 2 * static_cast<int>(s[1] * change_dis(gen)));
+ }
+ }
+
+ return shapes;
+ }
+
+ template <typename...>
+ void setup(TensorShape shape, DataType data_type)
+ {
+ // Generate input shapes
+ std::vector<TensorShape> src_shapes = generate_input_shapes(shape);
+
+ // Create tensors
+ _srcs.reserve(src_shapes.size());
+
+ std::vector<ITensorType *> src_ptrs;
+
+ for(const auto &shape : src_shapes)
+ {
+ _srcs.emplace_back(create_tensor<TensorType>(shape, data_type, 1, _fractional_bits));
+ src_ptrs.emplace_back(&_srcs.back());
+ }
+
+ TensorShape dst_shape = calculate_depth_concatenate_shape(src_ptrs);
+ _dst = create_tensor<TensorType>(dst_shape, data_type, 1, _fractional_bits);
+
+ _depth_concat.configure(src_ptrs, &_dst);
+
+ for(auto &src : _srcs)
+ {
+ src.allocator()->allocate();
+ }
+
+ _dst.allocator()->allocate();
+
+ int i = 0;
+ for(auto &src : _srcs)
+ {
+ library->fill_tensor_uniform(AccessorType(src), i++);
+ }
+ }
+
+ void run()
+ {
+ _depth_concat.run();
+ }
+
+ void teardown()
+ {
+ for(auto &src : _srcs)
+ {
+ src.allocator()->free();
+ }
+
+ _srcs.clear();
+
+ _dst.allocator()->free();
+ }
+
+private:
+ std::vector<TensorType> _srcs{};
+ TensorType _dst{};
+ Function _depth_concat{};
+ int _fractional_bits{ 1 };
+};
+} // namespace benchmark
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_DEPTHCONCATENATELAYERFIXTURE */