COMPMID-1602: Added Winograd tests for kernel 7x1 and 1x7.

Change-Id: I9250b2e8020fe87c6ed4de582bbc7460bbd8e94b
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/150287
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Isabella Gottardi <isabella.gottardi@arm.com>

5 files changed, 147 insertions, 1 deletions

diff --git a/tests/datasets/LargeConvolutionLayerDataset.h b/tests/datasets/LargeConvolutionLayerDataset.h
index ca38abb43a..40a8855ace 100644
--- a/tests/datasets/LargeConvolutionLayerDataset.h
+++ b/tests/datasets/LargeConvolutionLayerDataset.h
@@ -140,6 +140,42 @@ public:
     }
 };
 
+class LargeWinogradConvolutionLayer7x1Dataset final : public ConvolutionLayerDataset
+{
+public:
+    LargeWinogradConvolutionLayer7x1Dataset()
+    {
+        // Batch size 1
+        add_config(TensorShape(224U, 224U, 3U), TensorShape(7U, 1U, 3U, 64U), TensorShape(64U), TensorShape(218U, 224U, 64U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(123U, 134U, 16U), TensorShape(7U, 1U, 16U, 7U), TensorShape(7U), TensorShape(123U, 134U, 7U), PadStrideInfo(1, 1, 3, 0));
+        add_config(TensorShape(181U, 152U, 42U), TensorShape(7U, 1U, 42U, 100U), TensorShape(100U), TensorShape(175U, 152U, 100U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(200U, 201U, 24U), TensorShape(7U, 1U, 24U, 61), TensorShape(61U), TensorShape(200U, 201U, 61), PadStrideInfo(1, 1, 3, 0));
+
+        // Batch size 2, 3 and 4
+        add_config(TensorShape(224U, 224U, 3U, 2U), TensorShape(7U, 1U, 3U, 64U), TensorShape(64U), TensorShape(224U, 224U, 64U, 2U), PadStrideInfo(1, 1, 3, 0));
+        add_config(TensorShape(123U, 134U, 16U, 3U), TensorShape(7U, 1U, 16U, 7U), TensorShape(7U), TensorShape(123U, 134U, 7U, 3U), PadStrideInfo(1, 1, 3, 0));
+        add_config(TensorShape(181U, 152U, 42U, 4U), TensorShape(7U, 1U, 42U, 100U), TensorShape(100U), TensorShape(175U, 152U, 100U, 4U), PadStrideInfo(1, 1, 0, 0));
+    }
+};
+
+class LargeWinogradConvolutionLayer1x7Dataset final : public ConvolutionLayerDataset
+{
+public:
+    LargeWinogradConvolutionLayer1x7Dataset()
+    {
+        // Batch size 1
+        add_config(TensorShape(224U, 224U, 3U), TensorShape(1U, 7U, 3U, 64U), TensorShape(64U), TensorShape(224U, 218U, 64U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(123U, 134U, 16U), TensorShape(1U, 7U, 16U, 7U), TensorShape(7U), TensorShape(123U, 134U, 7U), PadStrideInfo(1, 1, 0, 3));
+        add_config(TensorShape(181U, 152U, 42U), TensorShape(1U, 7U, 42U, 100U), TensorShape(100U), TensorShape(181U, 146U, 100U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(200U, 201U, 24U), TensorShape(1U, 7U, 24U, 61), TensorShape(61U), TensorShape(200U, 201U, 61), PadStrideInfo(1, 1, 0, 3));
+
+        // Batch size 2, 3 and 4
+        add_config(TensorShape(224U, 224U, 3U, 2U), TensorShape(1U, 7U, 3U, 64U), TensorShape(64U), TensorShape(224U, 224U, 64U, 2U), PadStrideInfo(1, 1, 0, 3));
+        add_config(TensorShape(123U, 134U, 16U, 3U), TensorShape(1U, 7U, 16U, 7U), TensorShape(7U), TensorShape(123U, 134U, 7U, 3U), PadStrideInfo(1, 1, 0, 3));
+        add_config(TensorShape(181U, 152U, 42U, 4U), TensorShape(1U, 7U, 42U, 100U), TensorShape(100U), TensorShape(181U, 146U, 100U, 4U), PadStrideInfo(1, 1, 0, 0));
+    }
+};
+
 class LargeWinogradConvolutionLayer1x5Dataset final : public ConvolutionLayerDataset
 {
 public:
diff --git a/tests/datasets/SmallConvolutionLayerDataset.h b/tests/datasets/SmallConvolutionLayerDataset.h
index a288d07902..ca4abd1671 100644
--- a/tests/datasets/SmallConvolutionLayerDataset.h
+++ b/tests/datasets/SmallConvolutionLayerDataset.h
@@ -112,6 +112,26 @@ public:
     }
 };
 
+class SmallWinogradConvolutionLayer7x1Dataset final : public ConvolutionLayerDataset
+{
+public:
+    SmallWinogradConvolutionLayer7x1Dataset()
+    {
+        add_config(TensorShape(14U, 14U, 2U), TensorShape(7U, 1U, 2U, 1U), TensorShape(1U), TensorShape(8U, 14U, 1U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(14U, 14U, 2U), TensorShape(7U, 1U, 2U), TensorShape(1U), TensorShape(14U, 14U, 1U), PadStrideInfo(1, 1, 3, 0));
+    }
+};
+
+class SmallWinogradConvolutionLayer1x7Dataset final : public ConvolutionLayerDataset
+{
+public:
+    SmallWinogradConvolutionLayer1x7Dataset()
+    {
+        add_config(TensorShape(14U, 14U, 2U), TensorShape(1U, 7U, 2U, 1U), TensorShape(1U), TensorShape(14U, 8U, 1U), PadStrideInfo(1, 1, 0, 0));
+        add_config(TensorShape(14U, 14U, 2U), TensorShape(1U, 7U, 2U), TensorShape(1U), TensorShape(14U, 14U, 1U), PadStrideInfo(1, 1, 0, 3));
+    }
+};
+
 class SmallConvolutionLayerDataset final : public ConvolutionLayerDataset
 {
 public:
diff --git a/tests/validation/NEON/ConvolutionLayer.cpp b/tests/validation/NEON/ConvolutionLayer.cpp
index 61bf4a2c37..72e67a9acc 100644
--- a/tests/validation/NEON/ConvolutionLayer.cpp
+++ b/tests/validation/NEON/ConvolutionLayer.cpp
@@ -211,6 +211,50 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEWinogradConvolutionLayerFixture<float>, frame
 
 TEST_SUITE_END() // Conv5x1
 
+TEST_SUITE(Conv7x1)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEWinogradConvolutionLayerFixture<float>, framework::DatasetMode::PRECOMMIT,
+                       combine(combine(combine(datasets::SmallWinogradConvolutionLayer7x1Dataset(),
+                                               framework::dataset::make("DataType", { DataType::F32 })),
+                                       ActivationFunctionsDataset),
+                               framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, abs_tolerance_f32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEWinogradConvolutionLayerFixture<float>, framework::DatasetMode::NIGHTLY,
+                       combine(combine(combine(datasets::LargeWinogradConvolutionLayer7x1Dataset(),
+                                               framework::dataset::make("DataType", { DataType::F32 })),
+                                       ActivationFunctionsDataset),
+                               framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, abs_tolerance_1xN_f32);
+}
+TEST_SUITE_END() // Conv7x1
+
+TEST_SUITE(Conv1x7)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEWinogradConvolutionLayerFixture<float>, framework::DatasetMode::PRECOMMIT,
+                       combine(combine(combine(datasets::SmallWinogradConvolutionLayer1x7Dataset(),
+                                               framework::dataset::make("DataType", { DataType::F32 })),
+                                       ActivationFunctionsDataset),
+                               framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, abs_tolerance_f32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEWinogradConvolutionLayerFixture<float>, framework::DatasetMode::NIGHTLY,
+                       combine(combine(combine(datasets::LargeWinogradConvolutionLayer7x1Dataset(),
+                                               framework::dataset::make("DataType", { DataType::F32 })),
+                                       ActivationFunctionsDataset),
+                               framework::dataset::make("DataLayout", { DataLayout::NCHW, DataLayout::NHWC })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, abs_tolerance_1xN_f32);
+}
+TEST_SUITE_END() // Conv1x7
+
 TEST_SUITE(Conv3x3)
 FIXTURE_DATA_TEST_CASE(RunSmall, NEWinogradConvolutionLayerFixture<float>, framework::DatasetMode::PRECOMMIT,
                        combine(combine(combine(datasets::SmallWinogradConvolutionLayer3x3Dataset(),
diff --git a/tests/validation/fixtures/WinogradConvolutionLayerFixture.h b/tests/validation/fixtures/WinogradConvolutionLayerFixture.h
index 41f16d3a39..15ce201222 100644
--- a/tests/validation/fixtures/WinogradConvolutionLayerFixture.h
+++ b/tests/validation/fixtures/WinogradConvolutionLayerFixture.h
@@ -263,7 +263,17 @@ protected:
 
         // Set output tile
         Size2D output_tile(4U, 4U);
-        if(weights_shape[0] == 1)
+        if(weights_shape[0] == 7 && weights_shape[1] == 1)
+        {
+            output_tile.width  = 2;
+            output_tile.height = 1;
+        }
+        else if(weights_shape[0] == 1 && weights_shape[1] == 7)
+        {
+            output_tile.width  = 1;
+            output_tile.height = 2;
+        }
+        else if(weights_shape[0] == 1)
         {
             output_tile.width = 1;
         }
diff --git a/tests/validation/reference/Winograd.cpp b/tests/validation/reference/Winograd.cpp
index 3c2c11d632..294993b8d2 100644
--- a/tests/validation/reference/Winograd.cpp
+++ b/tests/validation/reference/Winograd.cpp
@@ -75,6 +75,18 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         0.f, -1.f, 0.f, 21.f / 4.f, 0.f, -21.f / 4.f, 0.f, 1.f
     };
 
+    static const float imatrix2x1_7x7[] =
+    {
+        -36.0f, 0.0f, 49.0f, 0.0f, -14.0f, 0.0f, 1.0f, 0.0f,
+        0.0f, -36.0f, 36.0f, 13.0f, -13.0f, -1.0f, 1.0f, 0.0f,
+        0.0f, 36.0f, 36.0f, -13.0f, -13.0f, 1.0f, 1.0f, 0.0f,
+        0.0f, -18.0f, 9.0f, 20.0f, -10.0f, -2.0f, 1.0f, 0.0f,
+        0.0f, 18.0f, 9.0f, -20.0f, -10.0f, 2.0f, 1.0f, 0.0f,
+        0.0f, -12.0f, 4.0f, 15.0f, -5.0f, -3.0f, 1.0f, 0.0f,
+        0.0f, 12.0f, 4.0f, -15.0f, -5.0f, 3.0f, 1.0f, 0.0f,
+        0.0f, -36.0f, 0.0f, 49.0f, 0.0f, -14.0f, 0.0f, 1.0f
+    };
+
     // ------------------------------------------
 
     // Winograd filter transform matrices
@@ -109,6 +121,18 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
 
     };
 
+    static const float fmatrix2x1_7x7[] =
+    {
+        -1.0f / 36.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f,
+        1.0f / 48.0f, -1.0f / 48.0f, 1.0f / 48.0f, -1.0f / 48.0f, 1.0f / 48.0f, -1.0f / 48.0f, 1.0f / 48.0f,
+        1.0f / 48.0f, 1.0f / 48.0f, 1.0f / 48.0f, 1.0f / 48.0f, 1.0f / 48.0f, 1.0f / 48.0f, 1.0f / 48.0f,
+        -1.0f / 120.0f, 1.0f / 60.0f, -1.0f / 30.0f, 1.0f / 15.0f, -2.0f / 15.0f, 4.0f / 15.0f, -8.0f / 15.0f,
+        -1.0f / 120.0f, -1.0f / 60.0f, -1.0f / 30.0f, -1.0f / 15.0f, -2.0f / 15.0f, -4.0f / 15.0f, -8.0f / 15.0f,
+        1.0f / 720.0f, -1.0f / 240.0f, 1.0f / 80.0f, -3.0f / 80.0f, 9.0f / 80.0f, -27.0f / 80.0f, 81.0f / 80.0f,
+        1.0f / 720.0f, 1.0f / 240.0f, 1.0f / 80.0f, 3.0f / 80.0f, 9.0f / 80.0f, 27.0f / 80.0f, 81.0f / 80.0f,
+        0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 0.0f, 1.0f
+    };
+
     // ------------------------------------------
 
     // Winograd output transform matrices
@@ -134,6 +158,12 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         0.0f, 1.0f, -1.0f, 8.0f, -8.0f, 1.0f, -1.0f, 1.0f
     };
 
+    static const float omatrix2x1_7x7[] =
+    {
+        1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 0.0f,
+        0.0f, -1.0f, 1.0f, -2.0f, 2.0f, -3.0f, 3.0f, 1.0f
+    };
+
     // ------------------------------------------
 
     using WinogradKey = std::tuple<std::pair<int, int>, std::pair<int, int>, WinogradTransformType>;
@@ -149,6 +179,8 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 3), WinogradTransformType::INPUT), imatrix4x4_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(5, 5), WinogradTransformType::INPUT), imatrix4x4_5x5 },
         { WinogradKey(std::pair<int, int>(4, 1), std::pair<int, int>(5, 1), WinogradTransformType::INPUT), imatrix4x4_5x5 },
+        { WinogradKey(std::pair<int, int>(2, 1), std::pair<int, int>(7, 1), WinogradTransformType::INPUT), imatrix2x1_7x7 },
+        { WinogradKey(std::pair<int, int>(1, 2), std::pair<int, int>(1, 7), WinogradTransformType::INPUT), imatrix2x1_7x7 },
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 5), WinogradTransformType::INPUT), imatrix4x4_5x5 },
         { WinogradKey(std::pair<int, int>(2, 2), std::pair<int, int>(3, 3), WinogradTransformType::FILTER), fmatrix2x2_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(3, 3), WinogradTransformType::FILTER), fmatrix4x4_3x3 },
@@ -158,6 +190,8 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 3), WinogradTransformType::FILTER), fmatrix4x4_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(5, 5), WinogradTransformType::FILTER), fmatrix4x4_5x5 },
         { WinogradKey(std::pair<int, int>(4, 1), std::pair<int, int>(5, 1), WinogradTransformType::FILTER), fmatrix4x4_5x5 },
+        { WinogradKey(std::pair<int, int>(2, 1), std::pair<int, int>(7, 1), WinogradTransformType::FILTER), fmatrix2x1_7x7 },
+        { WinogradKey(std::pair<int, int>(1, 2), std::pair<int, int>(1, 7), WinogradTransformType::FILTER), fmatrix2x1_7x7 },
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 5), WinogradTransformType::FILTER), fmatrix4x4_5x5 },
         { WinogradKey(std::pair<int, int>(2, 2), std::pair<int, int>(3, 3), WinogradTransformType::OUTPUT), omatrix2x2_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(3, 3), WinogradTransformType::OUTPUT), omatrix4x4_3x3 },
@@ -167,6 +201,8 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 3), WinogradTransformType::OUTPUT), omatrix4x4_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(5, 5), WinogradTransformType::OUTPUT), omatrix4x4_5x5 },
         { WinogradKey(std::pair<int, int>(4, 1), std::pair<int, int>(5, 1), WinogradTransformType::OUTPUT), omatrix4x4_5x5 },
+        { WinogradKey(std::pair<int, int>(2, 1), std::pair<int, int>(7, 1), WinogradTransformType::OUTPUT), omatrix2x1_7x7 },
+        { WinogradKey(std::pair<int, int>(1, 2), std::pair<int, int>(1, 7), WinogradTransformType::OUTPUT), omatrix2x1_7x7 },
         { WinogradKey(std::pair<int, int>(1, 4), std::pair<int, int>(1, 5), WinogradTransformType::OUTPUT), omatrix4x4_5x5 },
     };