Add Texture Pipe Support for Matmul Lhs T/NT Rhs NT kernels

Resolves: COMPMID-5945, COMPMID-5954

Change-Id: I7b27021d21f8e08c4896f6b1f595a75125064f9e
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9356
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 28 insertions, 0 deletions

diff --git a/tests/datasets/LargeMatMulDataset.h b/tests/datasets/LargeMatMulDataset.h
index cbc97d5e4a..b5181bc30b 100644
--- a/tests/datasets/LargeMatMulDataset.h
+++ b/tests/datasets/LargeMatMulDataset.h
@@ -54,6 +54,18 @@ public:
     }
 };
 
+class LargeMatMulDatasetRhsExportToCLImageRhsNT final : public MatMulDataset
+{
+public:
+    // For shape choices, please refer to the explanations given in SmallMatMulDatasetRhsExportToCLImageRhsNT
+    LargeMatMulDatasetRhsExportToCLImageRhsNT()
+    {
+        add_config(TensorShape(21U, 13U, 3U, 2U), TensorShape(32U, 21U, 3U, 2U), TensorShape(32U, 13U, 3U, 2U));
+        add_config(TensorShape(38U, 12U, 1U, 5U, 2U), TensorShape(20U, 38U, 1U, 5U, 2U), TensorShape(20U, 12U, 1U, 5U, 2U));
+        add_config(TensorShape(45U, 38U, 3U, 2U, 3U), TensorShape(20U, 45U, 3U, 2U, 3U), TensorShape(20U, 38U, 3U, 2U, 3U));
+    }
+};
+
 } // namespace datasets
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/datasets/SmallMatMulDataset.h b/tests/datasets/SmallMatMulDataset.h
index ae92b9abf5..93e5f7dc2c 100644
--- a/tests/datasets/SmallMatMulDataset.h
+++ b/tests/datasets/SmallMatMulDataset.h
@@ -57,6 +57,22 @@ public:
     }
 };
 
+class SmallMatMulDatasetRhsExportToCLImageRhsNT final : public MatMulDataset
+{
+public:
+    // Some considerations:
+    //  (1) N (Dimension 0 of Rhs matrix) dimension should be a multiple of 4
+    //  (2) Having N=20 enables us to test all possible N0 values, i.e. 4, 8, 16
+    //  (3) It's important to have more than one loop iterations in the K dimension
+    //      K has been chosen in accordance with K0
+    //  (4) The 5-th dimension has been chosen as non-unit because export_to_cl_iamge checks
+    //      were using dim1 * dim2 * dim3 to calculate the CLImage height; however, in our case
+    //      the tensor can be > 4D. To stress that case, the fifth dimension is chosen to be non-unit as well
+    SmallMatMulDatasetRhsExportToCLImageRhsNT()
+    {
+        add_config(TensorShape(7U, 3U, 2U, 1U, 2U), TensorShape(20U, 7U, 2U, 1U, 2U), TensorShape(20U, 3U, 2U, 1U, 2U));
+    }
+};
 } // namespace datasets
 } // namespace test
 } // namespace arm_compute