Add test for validating batched-GEMM on CPU

- Add new dataset for batched-GEMM
- Add test for running batched-GEMM without bias. Currently bias is not
  supported in batched-GEMM
- Fix reference implementation to slide correctly the RHS tensor

Resolves COMPMID-4588

Change-Id: I20fcb5d9160f44292b7cc34570add911b1d732f6
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6040
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/tests/datasets/SmallGEMMDataset.h b/tests/datasets/SmallGEMMDataset.h
index 7d2b42a..23b46a6 100644
--- a/tests/datasets/SmallGEMMDataset.h
+++ b/tests/datasets/SmallGEMMDataset.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 Arm Limited.
+ * Copyright (c) 2017-2021 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -50,6 +50,16 @@
         add_config(TensorShape(32U, 1U), TensorShape(17U, 32U), TensorShape(17U, 1U), TensorShape(17U, 1U), 0.4f, 0.7f);
     }
 };
+
+class SmallBatchedGEMMDataset final : public GEMMDataset
+{
+public:
+    SmallBatchedGEMMDataset()
+    {
+        add_config(TensorShape(2U, 4U, 1U, 3U), TensorShape(5U, 2U, 3U), TensorShape(5U), TensorShape(5U, 4U, 1U, 3U), 1.0f, 0.0f);
+    }
+};
+
 class SmallGEMMOutput3DDataset final : public GEMMDataset
 {
 public: