Fix Intermittent Neon™ ReduceMean QASYMM8 Mismatch

- Dividing scale by number of elements causes accuracy loss due to limitations in float datatype and truncation to int
- Adds rounding after division on aarch64 to negate this.

Resolves: [COMPMID-5839]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: I54ef0f7e56f39da1fa5f30378f551b5ca419a61d
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/492456
Tested-by: bsgcomp <bsgcomp@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9110
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>

1 files changed, 2 insertions, 1 deletions

diff --git a/tests/validation/reference/ReductionOperation.h b/tests/validation/reference/ReductionOperation.h
index 9c9e721b29..dd97778834 100644
--- a/tests/validation/reference/ReductionOperation.h
+++ b/tests/validation/reference/ReductionOperation.h
@@ -25,6 +25,7 @@
 #define ARM_COMPUTE_TEST_REDUCTION_OPERATION_H
 
 #include "tests/SimpleTensor.h"
+#include "arm_compute/core/Rounding.h"
 #include "tests/validation/Helpers.h"
 
 namespace arm_compute
@@ -37,7 +38,7 @@ namespace reference
 {
 template <typename T, typename OT>
 SimpleTensor<OT> reduction_operation(const SimpleTensor<T> &src, const TensorShape &dst_shape, unsigned int axis, ReductionOperation op,
-                                     QuantizationInfo quantization_info_output = QuantizationInfo());
+                                     QuantizationInfo quantization_info_output = QuantizationInfo(), RoundingPolicy policy = RoundingPolicy::TO_ZERO);
 } // namespace reference
 } // namespace validation
 } // namespace test