COMPMID-1037 Add support for F(4x4, 5x5) in CLWinogradInputTransformKernel

Change-Id: Iac26936f46d0f7cdd9d2f8393b0092cd5a223c45
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/127675
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>

3 files changed, 26 insertions, 2 deletions

diff --git a/tests/datasets/WinogradInputTransformDataset.h b/tests/datasets/WinogradInputTransformDataset.h
index 625daa0e6e..cbe63645de 100644
--- a/tests/datasets/WinogradInputTransformDataset.h
+++ b/tests/datasets/WinogradInputTransformDataset.h
@@ -109,6 +109,13 @@ public:
         add_config(TensorShape(27U, 13U, 2U, 4U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(27U, 13U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
         add_config(TensorShape(9U, 9U, 3U, 5U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(9U, 9U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
         add_config(TensorShape(14U, 14U, 512U, 2U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(14U, 14U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(9U, 9U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(9U, 9U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(27U, 13U, 2U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(27U, 13U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
+        add_config(TensorShape(128U, 64U, 1U, 3U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(128U, 64U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(9U, 9U, 3U, 4U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(9U, 9U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
+        add_config(TensorShape(27U, 13U, 2U, 4U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(27U, 13U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(9U, 9U, 3U, 5U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(9U, 9U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
+        add_config(TensorShape(14U, 14U, 512U, 2U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(14U, 14U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
     }
 };
 
@@ -121,6 +128,10 @@ public:
         add_config(TensorShape(57U, 60U, 13U, 8U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(57U, 60U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
         add_config(TensorShape(128U, 64U, 21U, 13U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(128U, 64U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
         add_config(TensorShape(83U, 72U, 14U, 5U), WinogradInfo(Size2D(2U, 2U), Size2D(3U, 3U), Size2D(83U, 72U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
+        add_config(TensorShape(42U, 37U, 8U, 15U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(42U, 37U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(57U, 60U, 13U, 8U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(57U, 60U), PadStrideInfo(1, 1, 1, 1), DataLayout::NCHW));
+        add_config(TensorShape(128U, 64U, 21U, 13U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(128U, 64U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
+        add_config(TensorShape(83U, 72U, 14U, 5U), WinogradInfo(Size2D(4U, 4U), Size2D(5U, 5U), Size2D(83U, 72U), PadStrideInfo(1, 1, 0, 0), DataLayout::NCHW));
     }
 };
 } // namespace datasets
diff --git a/tests/validation/CL/Winograd.cpp b/tests/validation/CL/Winograd.cpp
index 08810eaf58..4bfe4c8f42 100644
--- a/tests/validation/CL/Winograd.cpp
+++ b/tests/validation/CL/Winograd.cpp
@@ -126,14 +126,14 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLWinogradInputTransformFixture, framework::Dat
                                                                                                                      framework::dataset::make("DataLayout", { DataLayout::NCHW })),
                                                                                                              framework::dataset::make("DataType", { DataType::F32 })))
 {
-    validate(CLAccessor(_target), _reference);
+    validate(CLAccessor(_target), _reference, tolerance_f32);
 }
 
 FIXTURE_DATA_TEST_CASE(RunLarge, CLWinogradInputTransformFixture, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeWinogradInputTransformDataset(),
                                                                                                                    framework::dataset::make("DataLayout", { DataLayout::NCHW })),
                                                                                                            framework::dataset::make("DataType", { DataType::F32 })))
 {
-    validate(CLAccessor(_target), _reference);
+    validate(CLAccessor(_target), _reference, tolerance_f32);
 }
 TEST_SUITE_END() // InputTransform
 
diff --git a/tests/validation/reference/Winograd.cpp b/tests/validation/reference/Winograd.cpp
index 757a06d847..fa6e372cde 100644
--- a/tests/validation/reference/Winograd.cpp
+++ b/tests/validation/reference/Winograd.cpp
@@ -62,6 +62,18 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
         0.0f, 4.0f, 0.0f, -5.0f, 0.0f, 1.0f,
     };
 
+    static const float imatrix4x4_5x5[] =
+    {
+        1.f, 0.f, -21.f / 4.f, 0.f, 21.f / 4.f, 0.f, -1.f, 0.f,
+        0.f, 1.f, 1.f, -17.f / 4.f, -17.f / 4.f, 1.f, 1.f, 0.f,
+        0.f, -1.f, 1.f, 17.f / 4.f, -17.f / 4.f, -1.f, 1.f, 0.f,
+        0.f, 1.f / 2.f, 1.f / 4.f, -5.f / 2.f, -5.f / 4.f, 2.f, 1.f, 0.f,
+        0.f, -1.f / 2.f, 1.f / 4.f, 5.f / 2.f, -5.f / 4.f, -2.f, 1.f, 0.f,
+        0.f, 2.f, 4.f, -5.f / 2.f, -5.f, 1.f / 2.f, 1.f, 0.f,
+        0.f, -2.f, 4.f, 5.f / 2.f, -5.f, -1.f / 2.f, 1.f, 0.f,
+        0.f, -1.f, 0.f, 21.f / 4.f, 0.f, -21.f / 4.f, 0.f, 1.f
+    };
+
     // ------------------------------------------
 
     // Winograd filter transform matrices
@@ -122,6 +134,7 @@ void initialize_matrix_transform(SimpleTensor<T> &src, const Size2D &output_tile
     {
         { WinogradKey(std::pair<int, int>(2, 2), std::pair<int, int>(3, 3), WinogradTransformType::INPUT), imatrix2x2_3x3 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(3, 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>(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 },
         { WinogradKey(std::pair<int, int>(4, 4), std::pair<int, int>(5, 5), WinogradTransformType::FILTER), fmatrix4x4_5x5 },