Add optimization for global pooling in pooling_layer.cl

- Simplify the implementation when the pooling size has the same spatial
  dimensions of the input tensor
- Rework the heuristic for F32/F16
- Add test for validating the global pooling path
- Fix compare_dimensions in validation. The validation fails because we have different
  number of dimensions for NCHW and NHWC (e.g. 1,1,2,1(NCHW) -> 2,1,1,1(NHWC)

Change-Id: Iba680cb30bf2a5d0952265a4cc9794f368549ca5
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5510
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 105 insertions, 34 deletions

diff --git a/tests/validation/CL/PoolingLayer.cpp b/tests/validation/CL/PoolingLayer.cpp
index f42c187f8f..0153e659ae 100644
--- a/tests/validation/CL/PoolingLayer.cpp
+++ b/tests/validation/CL/PoolingLayer.cpp
@@ -159,12 +159,12 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLPoolingLayerFixture<float>, framework::Datase
     validate(CLAccessor(_target), _reference, tolerance_f32);
 }
 FIXTURE_DATA_TEST_CASE(RunMixedDataLayout, CLPoolingLayerMixedDataLayoutFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                        combine(combine(combine(combine(datasets::PoolingTypes(),
-                                                                                                        framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
-                                                                                                        framework::dataset::make("PadStride", { PadStrideInfo(2, 1, 0, 0) })),
-                                                                                                        framework::dataset::make("ExcludePadding", { false })),
-                                                                                                        framework::dataset::make("DataType", DataType::F32))),
-                                                                                                        pool_data_layout_dataset))
+                       combine(combine(combine(combine(datasets::PoolingTypes(),
+                                                       framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
+                                               framework::dataset::make("PadStride", { PadStrideInfo(2, 1, 0, 0) })),
+                                       framework::dataset::make("ExcludePadding", { false })),
+                               framework::dataset::make("DataType", DataType::F32))),
+                       pool_data_layout_dataset))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_f32);
@@ -188,6 +188,44 @@ FIXTURE_DATA_TEST_CASE(RunSmallIndices, CLPoolingLayerIndicesFixture<float>, fra
     validate(CLAccessor(_target_indices), _ref_indices);
 }
 
+TEST_SUITE(GlobalPooling)
+// *INDENT-OFF*
+// clang-format off
+FIXTURE_DATA_TEST_CASE(RunSmall, CLPoolingLayerFixture<float>, framework::DatasetMode::ALL,
+                       combine(combine(combine(combine(combine(combine(
+                                    framework::dataset::make("InputShape", { TensorShape(27U, 13U, 2U),
+                                                                             TensorShape(27U, 13U, 2U, 4U)
+                                                                           }),
+                                    framework::dataset::make("PoolingType", { PoolingType::AVG, PoolingType::L2, PoolingType::MAX })),
+                                    framework::dataset::make("PoolingSize", { Size2D(27, 13) })),
+                                    framework::dataset::make("PadStride", PadStrideInfo(1, 1, 0, 0))),
+                                    framework::dataset::make("ExcludePadding", false)),
+                                    framework::dataset::make("DataType", DataType::F32)),
+                                    framework::dataset::make("DataLayout", DataLayout::NHWC)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLPoolingLayerFixture<float>, framework::DatasetMode::NIGHTLY,
+                       combine(combine(combine(combine(combine(combine(
+                                    framework::dataset::make("InputShape", { TensorShape(79U, 37U, 11U),
+                                                                             TensorShape(79U, 37U, 11U, 4U)
+                                                                           }),
+                                    framework::dataset::make("PoolingType", { PoolingType::AVG, PoolingType::L2, PoolingType::MAX })),
+                                    framework::dataset::make("PoolingSize", { Size2D(79, 37) })),
+                                    framework::dataset::make("PadStride", PadStrideInfo(1, 1, 0, 0))),
+                                    framework::dataset::make("ExcludePadding", false)),
+                                    framework::dataset::make("DataType", DataType::F32)),
+                                    framework::dataset::make("DataLayout", DataLayout::NHWC)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f32);
+}
+// clang-format on
+// *INDENT-ON*
+TEST_SUITE_END() // GlobalPooling
+
 TEST_SUITE_END() // FP32
 
 TEST_SUITE(FP16)
@@ -216,6 +254,45 @@ FIXTURE_DATA_TEST_CASE(RunSmallIndices, CLPoolingLayerIndicesFixture<half>, fram
     validate(CLAccessor(_target), _reference, tolerance_f32);
     validate(CLAccessor(_target_indices), _ref_indices);
 }
+
+TEST_SUITE(GlobalPooling)
+// *INDENT-OFF*
+// clang-format off
+FIXTURE_DATA_TEST_CASE(RunSmall, CLPoolingLayerFixture<half>, framework::DatasetMode::ALL,
+                       combine(combine(combine(combine(combine(combine(
+                                    framework::dataset::make("InputShape", { TensorShape(27U, 13U, 2U),
+                                                                             TensorShape(27U, 13U, 2U, 4U)
+                                                                            }),
+                                    framework::dataset::make("PoolingType", { PoolingType::AVG, PoolingType::L2, PoolingType::MAX })),
+                                    framework::dataset::make("PoolingSize", { Size2D(27, 13) })),
+                                    framework::dataset::make("PadStride", PadStrideInfo(1, 1, 0, 0))),
+                                    framework::dataset::make("ExcludePadding", false)),
+                                    framework::dataset::make("DataType", DataType::F16)),
+                                    framework::dataset::make("DataLayout", DataLayout::NHWC)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f16);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLPoolingLayerFixture<half>, framework::DatasetMode::NIGHTLY,
+                       combine(combine(combine(combine(combine(combine(
+                                    framework::dataset::make("InputShape", { TensorShape(79U, 37U, 11U),
+                                                                             TensorShape(79U, 37U, 11U, 4U)
+                                                                           }),
+                                    framework::dataset::make("PoolingType", { PoolingType::AVG, PoolingType::L2, PoolingType::MAX })),
+                                    framework::dataset::make("PoolingSize", { Size2D(79, 37) })),
+                                    framework::dataset::make("PadStride", PadStrideInfo(1, 1, 0, 0))),
+                                    framework::dataset::make("ExcludePadding", false)),
+                                    framework::dataset::make("DataType", DataType::F16)),
+                                    framework::dataset::make("DataLayout", DataLayout::NHWC)))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_f16);
+}
+// clang-format on
+// *INDENT-ON*
+TEST_SUITE_END() // GlobalPooling
+
 TEST_SUITE_END() // FP16
 TEST_SUITE_END() // Float
 
@@ -238,14 +315,14 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLPoolingLayerQuantizedFixture<uint8_t>, framew
     validate(CLAccessor(_target), _reference, tolerance_qasymm8);
 }
 FIXTURE_DATA_TEST_CASE(RunMixedDataLayout, CLPoolingLayerQuantizedMixedDataLayoutFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
-                                                                                                                    combine(combine(combine(combine(framework::dataset::make("PoolingType", { PoolingType::MAX, PoolingType::AVG }),
-                                                                                                                    framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
-                                                                                                                    framework::dataset::make("PadStride", { PadStrideInfo(1, 2, 1, 1) })),
-                                                                                                                    framework::dataset::make("ExcludePadding", { true })),
-                                                                                                                    framework::dataset::make("DataType", DataType::QASYMM8))),
-                                                                                                                    framework::dataset::make("DataLayout", { DataLayout::NHWC, DataLayout::NCHW })),
-                                                                                                                    framework::dataset::make("InputQuantInfo", { QuantizationInfo(1.f / 255.f, 10) })),
-                                                                                                                    framework::dataset::make("OutputQuantInfo", { QuantizationInfo(1.f / 255.f, 5) })))
+                       combine(combine(combine(combine(framework::dataset::make("PoolingType", { PoolingType::MAX, PoolingType::AVG }),
+                                                       framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
+                                               framework::dataset::make("PadStride", { PadStrideInfo(1, 2, 1, 1) })),
+                                       framework::dataset::make("ExcludePadding", { true })),
+                               framework::dataset::make("DataType", DataType::QASYMM8))),
+                       framework::dataset::make("DataLayout", { DataLayout::NHWC, DataLayout::NCHW })),
+                       framework::dataset::make("InputQuantInfo", { QuantizationInfo(1.f / 255.f, 10) })),
+                       framework::dataset::make("OutputQuantInfo", { QuantizationInfo(1.f / 255.f, 5) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_qasymm8);
@@ -264,14 +341,14 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLPoolingLayerQuantizedFixture<int8_t>, framewo
     validate(CLAccessor(_target), _reference, tolerance_qasymm8_s);
 }
 FIXTURE_DATA_TEST_CASE(RunMixedDataLayout, CLPoolingLayerQuantizedMixedDataLayoutFixture<int8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
-                                                                                                                    combine(combine(combine(combine(framework::dataset::make("PoolingType", { PoolingType::MAX, PoolingType::AVG }),
-                                                                                                                    framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
-                                                                                                                    framework::dataset::make("PadStride", { PadStrideInfo(1, 2, 1, 1) })),
-                                                                                                                    framework::dataset::make("ExcludePadding", { true })),
-                                                                                                                    framework::dataset::make("DataType", DataType::QASYMM8_SIGNED))),
-                                                                                                                    framework::dataset::make("DataLayout", { DataLayout::NHWC, DataLayout::NCHW })),
-                                                                                                                    framework::dataset::make("InputQuantInfo", { QuantizationInfo(1.f / 127.f, -10) })),
-                                                                                                                    framework::dataset::make("OutputQuantInfo", { QuantizationInfo(1.f / 127.f, -10) })))
+                       combine(combine(combine(combine(framework::dataset::make("PoolingType", { PoolingType::MAX, PoolingType::AVG }),
+                                                       framework::dataset::make("PoolingSize", { Size2D(2, 2) })),
+                                               framework::dataset::make("PadStride", { PadStrideInfo(1, 2, 1, 1) })),
+                                       framework::dataset::make("ExcludePadding", { true })),
+                               framework::dataset::make("DataType", DataType::QASYMM8_SIGNED))),
+                       framework::dataset::make("DataLayout", { DataLayout::NHWC, DataLayout::NCHW })),
+                       framework::dataset::make("InputQuantInfo", { QuantizationInfo(1.f / 127.f, -10) })),
+                       framework::dataset::make("OutputQuantInfo", { QuantizationInfo(1.f / 127.f, -10) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_qasymm8_s);
diff --git a/tests/validation/Validation.h b/tests/validation/Validation.h
index a75562bac2..f1ce0fecc7 100644
--- a/tests/validation/Validation.h
+++ b/tests/validation/Validation.h
@@ -159,11 +159,13 @@ bool compare_dimensions(const Dimensions<T> &dimensions1, const Dimensions<T> &d
     {
         // In case a 1D/2D shape becomes 3D after permutation, the permuted tensor will have two/one dimension(s) more and the first (two) value(s) will be 1
         // clang-format off
-        if((dimensions1.num_dimensions() != dimensions2.num_dimensions()) &&
-           ((dimensions1.num_dimensions() != (dimensions2.num_dimensions() + 1)) || (dimensions1.x() != 1)) &&
-           ((dimensions1.num_dimensions() != (dimensions2.num_dimensions() + 2)) || (dimensions1.x() != 1) || (dimensions1.y() != 1)))
+        const auto max_dims = std::max(dimensions1.num_dimensions(), dimensions2.num_dimensions());
+        for(unsigned int i = 3; i < max_dims; ++i)
         {
-            return false;
+            if(dimensions1[i] != dimensions2[i])
+            {
+                return false;
+            }
         }
         // clang-format on
 
@@ -171,14 +173,6 @@ bool compare_dimensions(const Dimensions<T> &dimensions1, const Dimensions<T> &d
         {
             return false;
         }
-
-        for(unsigned int i = 3; i < dimensions1.num_dimensions(); ++i)
-        {
-            if(dimensions1[i] != dimensions2[i])
-            {
-                return false;
-            }
-        }
     }
 
     return true;