COMPMID-1816: Use parallel reduction on 0 axis in CL ARG_MIN/ARG_MAX

Parallelization of reduction along x axes
Removal of the use of padding
Fast vector implementation of reduction operation

Change-Id: I3a56c57b9fc1135cf8f79d1021d966ea22b084b1
Signed-off-by: Manuel Bottini <manuel.bottini@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1791
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

2 files changed, 140 insertions, 0 deletions

diff --git a/tests/benchmark/CL/ArgMinMax.cpp b/tests/benchmark/CL/ArgMinMax.cpp
new file mode 100644
index 0000000000..25a4a05d44
--- /dev/null
+++ b/tests/benchmark/CL/ArgMinMax.cpp
@@ -0,0 +1,56 @@
+/*
+ * 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/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLArgMinMaxLayer.h"
+
+#include "tests/CL/CLAccessor.h"
+#include "tests/benchmark/fixtures/ArgMinMaxFixture.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/datasets/SplitDataset.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace benchmark
+{
+TEST_SUITE(CL)
+
+using CLArgMinMaxBenchmarkFixture = ArgMinMaxBenchmarkFixture<CLTensor, CLArgMinMaxLayer, CLAccessor>;
+
+REGISTER_FIXTURE_DATA_TEST_CASE(ArgMinMax, CLArgMinMaxBenchmarkFixture, framework::DatasetMode::PRECOMMIT,
+                                framework::dataset::combine(framework::dataset::combine(framework::dataset::combine(
+                                                                                            datasets::Large3DShapes(),
+                                                                                            framework::dataset::make("DataType", { DataType::F32 })),
+                                                                                        framework::dataset::make("Axis", { 0, 1 })),
+                                                            framework::dataset::make("ReductionOperation", { ReductionOperation::ARG_IDX_MAX, ReductionOperation::ARG_IDX_MIN })));
+
+TEST_SUITE_END() // CL
+} // namespace benchmark
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/benchmark/fixtures/ArgMinMaxFixture.h b/tests/benchmark/fixtures/ArgMinMaxFixture.h
new file mode 100644
index 0000000000..f1a0c5ab4b
--- /dev/null
+++ b/tests/benchmark/fixtures/ArgMinMaxFixture.h
@@ -0,0 +1,84 @@
+/*
+ * 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.
+ */
+#ifndef ARM_COMPUTE_TEST_ARGMINMAXFIXTURE
+#define ARM_COMPUTE_TEST_ARGMINMAXFIXTURE
+
+#include "arm_compute/core/TensorShape.h"
+#include "arm_compute/core/Types.h"
+#include "tests/Globals.h"
+#include "tests/Utils.h"
+#include "tests/framework/Fixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace benchmark
+{
+/** Fixture that can be used for NEON and CL */
+template <typename TensorType, typename Function, typename Accessor>
+class ArgMinMaxBenchmarkFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(TensorShape shape, DataType data_type, int axis, ReductionOperation op)
+    {
+        // Create tensors
+        src = create_tensor<TensorType>(shape, data_type);
+        dst = create_tensor<TensorType>(shape, DataType::U32);
+
+        // Create and configure function
+        argminmax_layer.configure(&src, axis, &dst, op);
+
+        // Allocate tensors
+        src.allocator()->allocate();
+        dst.allocator()->allocate();
+    }
+
+    void run()
+    {
+        argminmax_layer.run();
+    }
+
+    void sync()
+    {
+        sync_if_necessary<TensorType>();
+        sync_tensor_if_necessary<TensorType>(dst);
+    }
+
+    void teardown()
+    {
+        src.allocator()->free();
+        dst.allocator()->free();
+    }
+
+private:
+    TensorType src{};
+    TensorType dst{};
+    Function   argminmax_layer{};
+};
+} // namespace benchmark
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_ARGMINMAXFIXTURE */