Add in place summation to CPU GEMM kernels

Instead of dispatching the sum postop for GEMM kernels to a
separate kernel + add, that requires an extra destination sized
allocation, plus 3 extra load/stores per element,
just do it in the GEMM kernel.

Resolves: ONCPUML-1442

Signed-off-by: Radu Salavat <radu.salavat@arm.com>
Co-authored-by: Milos Puzovic <milos.puzovic@arm.com>
Change-Id: I7a1f2da3300875fa1ac88b705a34390969518077
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11298
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

2 files changed, 32 insertions, 8 deletions

diff --git a/tests/datasets/LargeGEMMDataset.h b/tests/datasets/LargeGEMMDataset.h
index 6cdff7f559..e45319ef57 100644
--- a/tests/datasets/LargeGEMMDataset.h
+++ b/tests/datasets/LargeGEMMDataset.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 Arm Limited.
+ * Copyright (c) 2017-2019, 2024 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef ARM_COMPUTE_TEST_LARGE_GEMM_DATASET
-#define ARM_COMPUTE_TEST_LARGE_GEMM_DATASET
+#ifndef ACL_TESTS_DATASETS_LARGEGEMMDATASET_H
+#define ACL_TESTS_DATASETS_LARGEGEMMDATASET_H
 
 #include "tests/datasets/GEMMDataset.h"
 
@@ -79,7 +79,20 @@ public:
         add_config(TensorShape(1729U, 17U, 10U, 3U), TensorShape(128U, 1729U), TensorShape(128U), TensorShape(128U, 17U, 10U, 3U), 1.0f, 0.3f);
     }
 };
+
+class LargeAccumulateGEMMDataset final : public GEMMDataset
+{
+public:
+    LargeAccumulateGEMMDataset()
+    {
+        add_config(TensorShape(923U, 429U), TensorShape(871U, 923U), TensorShape(871U, 429U), TensorShape(871U, 429U), 1.0f, 0.0f);
+        add_config(TensorShape(1021U, 1U), TensorShape(783U, 1021U), TensorShape(783U, 1U), TensorShape(783U, 1U), 1.0f, 0.0f);
+        add_config(TensorShape(1021U, 1U), TensorShape(783U, 1021U), TensorShape(783U, 1U), TensorShape(783U, 1U), 1.0f, 0.0f);
+        add_config(TensorShape(941U, 1U), TensorShape(623U, 941U), TensorShape(623U, 1U), TensorShape(623U, 1U), 1.0f, 0.0f);
+    }
+};
+
 } // namespace datasets
 } // namespace test
 } // namespace arm_compute
-#endif /* ARM_COMPUTE_TEST_LARGE_GEMM_DATASET */
+#endif // ACL_TESTS_DATASETS_LARGEGEMMDATASET_H
diff --git a/tests/datasets/SmallGEMMDataset.h b/tests/datasets/SmallGEMMDataset.h
index c12f57b266..99c7abbf64 100644
--- a/tests/datasets/SmallGEMMDataset.h
+++ b/tests/datasets/SmallGEMMDataset.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2023 Arm Limited.
+ * Copyright (c) 2017-2024 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -21,8 +21,8 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#ifndef ARM_COMPUTE_TEST_SMALL_GEMM_DATASET
-#define ARM_COMPUTE_TEST_SMALL_GEMM_DATASET
+#ifndef ACL_TESTS_DATASETS_SMALLGEMMDATASET_H
+#define ACL_TESTS_DATASETS_SMALLGEMMDATASET_H
 
 #include "tests/datasets/GEMMDataset.h"
 
@@ -97,7 +97,18 @@ public:
     }
 };
 
+class SmallAccumulateGEMMDataset final : public GEMMDataset
+{
+public:
+    SmallAccumulateGEMMDataset()
+    {
+        add_config(TensorShape(8U, 2U), TensorShape(16U, 8U), TensorShape(16U, 2U), TensorShape(16U, 2U), 1.0f, 0.0f);
+        add_config(TensorShape(31U, 1U), TensorShape(23U, 31U), TensorShape(23U, 1U), TensorShape(23U, 1U), 1.0f, 0.0f);
+        add_config(TensorShape(21U, 13U), TensorShape(33U, 21U), TensorShape(33U, 13U), TensorShape(33U, 13U), 1.0f, 0.0f);
+    }
+};
+
 } // namespace datasets
 } // namespace test
 } // namespace arm_compute
-#endif /* ARM_COMPUTE_TEST_SMALL_GEMM_DATASET */
+#endif // ACL_TESTS_DATASETS_SMALLGEMMDATASET_H