COMPMID-1566: Add broadcast to CLArithmeticSubtraction

Change-Id: I05d21f9a92013ecfd1128d12cf1561cfd6e5c5e9
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/147983
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

15 files changed, 286 insertions, 507 deletions

diff --git a/tests/validation/CL/ArithmeticAddition.cpp b/tests/validation/CL/ArithmeticAddition.cpp
index bd27bc4043..e327769397 100644
--- a/tests/validation/CL/ArithmeticAddition.cpp
+++ b/tests/validation/CL/ArithmeticAddition.cpp
@@ -33,7 +33,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/ArithmeticAdditionFixture.h"
+#include "tests/validation/fixtures/ArithmeticOperationsFixture.h"
 
 namespace arm_compute
 {
diff --git a/tests/validation/CL/ArithmeticSubtraction.cpp b/tests/validation/CL/ArithmeticSubtraction.cpp
index b19d963515..cd13f42ec4 100644
--- a/tests/validation/CL/ArithmeticSubtraction.cpp
+++ b/tests/validation/CL/ArithmeticSubtraction.cpp
@@ -18,7 +18,7 @@
  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONCLCTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
 #include "arm_compute/core/Types.h"
@@ -33,7 +33,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/ArithmeticSubtractionFixture.h"
+#include "tests/validation/fixtures/ArithmeticOperationsFixture.h"
 
 namespace arm_compute
 {
@@ -47,14 +47,11 @@ namespace
 const auto ArithmeticSubtractionU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)),
                                                     framework::dataset::make("DataType",
                                                                              DataType::U8));
-const auto ArithmeticSubtractionS16Dataset = combine(combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::S16)),
+const auto ArithmeticSubtractionQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                         framework::dataset::make("DataType",
+                                                                                  DataType::QASYMM8));
+const auto ArithmeticSubtractionS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
                                                      framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionU8U8S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)),
-                                                         framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionS16U8S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::U8)),
-                                                          framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionU8S16S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S16)),
-                                                          framework::dataset::make("DataType", DataType::S16));
 const auto ArithmeticSubtractionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
                                                       framework::dataset::make("DataType", DataType::F16));
 const auto ArithmeticSubtractionFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
@@ -67,34 +64,34 @@ TEST_SUITE(ArithmeticSubtraction)
 // *INDENT-OFF*
 // clang-format off
 DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
-               framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),      // Window shrink
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Invalid data type combination
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),     // Mismatching shapes
-                                                      }),
-               framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
-                                                       TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
-                                                     })),
-               framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
-                                                     })),
-               framework::dataset::make("Expected", { true, true, false, false, false, false, true })),
-               input1_info, input2_info, output_info, expected)
+        framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                 TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),      // Window shrink
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Invalid data type combination
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),     // Mismatching shapes
+        }),
+        framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
+                                                TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
+        })),
+        framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
+        })),
+        framework::dataset::make("Expected", { true, true, false, false, false})),
+        input1_info, input2_info, output_info, expected)
 {
     ARM_COMPUTE_EXPECT(bool(CLArithmeticSubtraction::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), ConvertPolicy::WRAP)) == expected, framework::LogLevel::ERRORS);
 }
 // clang-format on
 // *INDENT-ON*
 
-template <typename T1, typename T2 = T1, typename T3 = T1>
-using CLArithmeticSubtractionFixture = ArithmeticSubtractionValidationFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T1, T2, T3>;
+template <typename T>
+using CLArithmeticSubtractionFixture = ArithmeticSubtractionValidationFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
 
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
@@ -120,33 +117,30 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::da
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                                     ArithmeticSubtractionU8Dataset),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionU8Dataset),
                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // U8
 
-template <typename T1, typename T2 = T1>
-using CLArithmeticSubtractionToS16Fixture = CLArithmeticSubtractionFixture<T1, T2, int16_t>;
+template <typename T>
+using CLArithmeticSubtractionQuantizedFixture = ArithmeticSubtractionValidationQuantizedFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
 
-TEST_SUITE(S16)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
-                                                                                   framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-                                                                           framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
-               shape, data_type1, data_type2, policy)
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+               shape, policy)
 {
     // Create tensors
-    CLTensor ref_src1 = create_tensor<CLTensor>(shape, data_type1);
-    CLTensor ref_src2 = create_tensor<CLTensor>(shape, data_type2);
-    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::S16);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QASYMM8);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QASYMM8);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8);
 
     // Create and Configure function
-    CLArithmeticSubtraction sub;
-    sub.configure(&ref_src1, &ref_src2, &dst, policy);
+    CLArithmeticSubtraction add;
+    add.configure(&ref_src1, &ref_src2, &dst, policy);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -158,78 +152,61 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combi
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
 }
-TEST_SUITE(S16_S16_S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionToS16Fixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
 
-FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionToS16Fixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
-                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
-TEST_SUITE_END()
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
+                       ArithmeticSubtractionQASYMM8Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       framework::dataset::make("QuantizationInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                       framework::dataset::make("QuantizationInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                       framework::dataset::make("QuantizationInfo", { QuantizationInfo(1.f / 255.f, 5) }))
 
-TEST_SUITE(U8_U8_S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionToS16Fixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                       ArithmeticSubtractionU8U8S16Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+                      )
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
+TEST_SUITE_END() // QASYMM8
+TEST_SUITE_END() // Quantized
 
-FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionToS16Fixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                                        ArithmeticSubtractionU8U8S16Dataset),
-                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+               shape, data_type, policy)
 {
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
-TEST_SUITE_END()
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, data_type);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::S16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::S16);
 
-TEST_SUITE(S16_U8_S16)
-using CLAriSubS16U8ToS16Fixture = CLArithmeticSubtractionToS16Fixture<int16_t, uint8_t>;
-FIXTURE_DATA_TEST_CASE(RunSmall, CLAriSubS16U8ToS16Fixture, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                               ArithmeticSubtractionS16U8S16Dataset),
-                                                                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
+    // Create and Configure function
+    CLArithmeticSubtraction sub;
+    sub.configure(&ref_src1, &ref_src2, &dst, policy);
 
-FIXTURE_DATA_TEST_CASE(RunLarge, CLAriSubS16U8ToS16Fixture, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                             ArithmeticSubtractionS16U8S16Dataset),
-                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
 }
-TEST_SUITE_END()
 
-TEST_SUITE(U8_S16_S16)
-using CLAriSubU8S16ToS16Fixture = CLArithmeticSubtractionToS16Fixture<uint8_t, int16_t>;
-FIXTURE_DATA_TEST_CASE(RunSmall, CLAriSubU8S16ToS16Fixture, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                               ArithmeticSubtractionU8S16S16Dataset),
-                                                                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, CLAriSubU8S16ToS16Fixture, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                             ArithmeticSubtractionU8S16S16Dataset),
-                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // S16
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -239,7 +216,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<half>, framework
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // FP16
 
 TEST_SUITE(FP32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
@@ -278,11 +255,30 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionFixture<float>, framewor
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+template <typename T>
+using CLArithmeticSubtractionBroadcastFixture = ArithmeticSubtractionBroadcastValidationFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLArithmeticSubtractionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapesBroadcast(),
+                       ArithmeticSubtractionFP32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLArithmeticSubtractionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapesBroadcast(),
+                       ArithmeticSubtractionFP32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // FP32
+TEST_SUITE_END() // Float
+
+TEST_SUITE_END() // ArithmeticSubtraction
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
-} // namespace arm_compute
+} // namespace arm_compute
\ No newline at end of file
diff --git a/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp b/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
index b16503e079..82946fa209 100755
--- a/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
+++ b/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -33,7 +33,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/ArithmeticAdditionFixture.h"
+#include "tests/validation/fixtures/ArithmeticOperationsFixture.h"
 
 namespace arm_compute
 {
diff --git a/tests/validation/NEON/ArithmeticAddition.cpp b/tests/validation/NEON/ArithmeticAddition.cpp
index 3632c3c207..4dc56dce54 100644
--- a/tests/validation/NEON/ArithmeticAddition.cpp
+++ b/tests/validation/NEON/ArithmeticAddition.cpp
@@ -33,7 +33,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/ArithmeticAdditionFixture.h"
+#include "tests/validation/fixtures/ArithmeticOperationsFixture.h"
 
 namespace arm_compute
 {
diff --git a/tests/validation/NEON/ArithmeticSubtraction.cpp b/tests/validation/NEON/ArithmeticSubtraction.cpp
index 210ed4578f..2e21c14ebc 100644
--- a/tests/validation/NEON/ArithmeticSubtraction.cpp
+++ b/tests/validation/NEON/ArithmeticSubtraction.cpp
@@ -33,7 +33,7 @@
 #include "tests/framework/Macros.h"
 #include "tests/framework/datasets/Datasets.h"
 #include "tests/validation/Validation.h"
-#include "tests/validation/fixtures/ArithmeticSubtractionFixture.h"
+#include "tests/validation/fixtures/ArithmeticOperationsFixture.h"
 
 namespace arm_compute
 {
@@ -49,12 +49,6 @@ const auto ArithmeticSubtractionU8Dataset = combine(combine(framework::dataset::
                                                                              DataType::U8));
 const auto ArithmeticSubtractionS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
                                                      framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionU8U8S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)),
-                                                         framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionS16U8S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::U8)),
-                                                          framework::dataset::make("DataType", DataType::S16));
-const auto ArithmeticSubtractionU8S16S16Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S16)),
-                                                          framework::dataset::make("DataType", DataType::S16));
 #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 const auto ArithmeticSubtractionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
                                                       framework::dataset::make("DataType", DataType::F16));
@@ -66,38 +60,38 @@ const auto ArithmeticSubtractionFP32Dataset = combine(combine(framework::dataset
 TEST_SUITE(NEON)
 TEST_SUITE(ArithmeticSubtraction)
 
+template <typename T>
+using NEArithmeticSubtractionFixture = ArithmeticSubtractionValidationFixture<Tensor, Accessor, NEArithmeticSubtraction, T>;
+
 // *INDENT-OFF*
 // clang-format off
 DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
-               framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),      // Window shrink
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Invalid data type combination
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),     // Mismatching shapes
-                                                      }),
-               framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
-                                                       TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
-                                                     })),
-               framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
-                                                     })),
-               framework::dataset::make("Expected", { true, true, false, false, false})),
-               input1_info, input2_info, output_info, expected)
+        framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                 TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),      // Window shrink
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),      // Invalid data type combination
+                                                 TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),     // Mismatching shapes
+        }),
+        framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
+                                                TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
+        })),
+        framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
+        })),
+        framework::dataset::make("Expected", { true, true, false, false, false})),
+        input1_info, input2_info, output_info, expected)
 {
     ARM_COMPUTE_EXPECT(bool(NEArithmeticSubtraction::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), ConvertPolicy::WRAP)) == expected, framework::LogLevel::ERRORS);
 }
 // clang-format on
 // *INDENT-ON*
 
-template <typename T1, typename T2 = T1, typename T3 = T1>
-using NEArithmeticSubtractionFixture = ArithmeticSubtractionValidationFixture<Tensor, Accessor, NEArithmeticSubtraction, T1, T2, T3>;
-
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                shape, policy)
@@ -128,21 +122,16 @@ FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticSubtractionFixture<uint8_t>, framew
     // Validate output
     validate(Accessor(_target), _reference);
 }
-TEST_SUITE_END()
-
-template <typename T1, typename T2 = T1>
-using NEArithmeticSubtractionToS16Fixture = NEArithmeticSubtractionFixture<T1, T2, int16_t>;
+TEST_SUITE_END() // U8
 
 TEST_SUITE(S16)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
-                                                                                   framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-                                                                           framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
-               shape, data_type1, data_type2, policy)
+               shape, data_type, policy)
 {
     // Create tensors
-    Tensor ref_src1 = create_tensor<Tensor>(shape, data_type1);
-    Tensor ref_src2 = create_tensor<Tensor>(shape, data_type2);
+    Tensor ref_src1 = create_tensor<Tensor>(shape, data_type);
+    Tensor ref_src2 = create_tensor<Tensor>(shape, DataType::S16);
     Tensor dst      = create_tensor<Tensor>(shape, DataType::S16);
 
     // Create and Configure function
@@ -160,92 +149,34 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(combi
     validate(dst.info()->padding(), padding);
 }
 
-TEST_SUITE(S16_S16_S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticSubtractionToS16Fixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticSubtractionToS16Fixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
-                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-TEST_SUITE_END()
-
-TEST_SUITE(U8_U8_S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticSubtractionToS16Fixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                       ArithmeticSubtractionU8U8S16Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticSubtractionToS16Fixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                                        ArithmeticSubtractionU8U8S16Dataset),
-                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-TEST_SUITE_END()
-
-TEST_SUITE(S16_U8_S16)
-using NEAriSubS16U8ToS16Fixture = NEArithmeticSubtractionToS16Fixture<int16_t, uint8_t>;
-FIXTURE_DATA_TEST_CASE(RunSmall, NEAriSubS16U8ToS16Fixture, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                               ArithmeticSubtractionS16U8S16Dataset),
-                                                                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-
-FIXTURE_DATA_TEST_CASE(RunLarge, NEAriSubS16U8ToS16Fixture, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                             ArithmeticSubtractionS16U8S16Dataset),
-                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(Accessor(_target), _reference);
-}
-TEST_SUITE_END()
-
-TEST_SUITE(U8_S16_S16)
-using NEAriSubU8S16ToS16Fixture = NEArithmeticSubtractionToS16Fixture<uint8_t, int16_t>;
-FIXTURE_DATA_TEST_CASE(RunSmall, NEAriSubU8S16ToS16Fixture, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(),
-                                                                                                               ArithmeticSubtractionU8S16S16Dataset),
-                                                                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, NEAriSubU8S16ToS16Fixture, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(),
-                                                                                                             ArithmeticSubtractionU8S16S16Dataset),
-                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // S16
 
 TEST_SUITE(Float)
 #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
-TEST_SUITE(FP16)
+TEST_SUITE(F16)
 FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticSubtractionFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionFP16Dataset),
                                                                                                             framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
-TEST_SUITE_END()
-#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+TEST_SUITE_END() // F16
+#endif           /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
 
-TEST_SUITE(FP32)
+TEST_SUITE(F32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                shape, policy)
 {
@@ -282,11 +213,30 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticSubtractionFixture<float>, framewor
     // Validate output
     validate(Accessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+template <typename T>
+using NEArithmeticSubtractionBroadcastFixture = ArithmeticSubtractionBroadcastValidationFixture<Tensor, Accessor, NEArithmeticSubtraction, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticSubtractionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapesBroadcast(),
+                       ArithmeticSubtractionFP32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, NEArithmeticSubtractionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapesBroadcast(),
+                       ArithmeticSubtractionFP32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // F32
+TEST_SUITE_END() // Float
+
+TEST_SUITE_END() // ArithmeticSubtraction
+TEST_SUITE_END() // NEON
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/fixtures/ArithmeticAdditionFixture.h b/tests/validation/fixtures/ArithmeticOperationsFixture.h
index 66875934de..fb46a5185e 100644
--- a/tests/validation/fixtures/ArithmeticAdditionFixture.h
+++ b/tests/validation/fixtures/ArithmeticOperationsFixture.h
@@ -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_ARITHMETIC_ADDITION_FIXTURE
-#define ARM_COMPUTE_TEST_ARITHMETIC_ADDITION_FIXTURE
+#ifndef ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_FIXTURE
+#define ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_FIXTURE
 
 #include "arm_compute/core/TensorShape.h"
 #include "arm_compute/core/Types.h"
@@ -32,7 +32,7 @@
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Fixture.h"
 #include "tests/validation/Helpers.h"
-#include "tests/validation/reference/ArithmeticAddition.h"
+#include "tests/validation/reference/ArithmeticOperations.h"
 
 namespace arm_compute
 {
@@ -41,13 +41,15 @@ namespace test
 namespace validation
 {
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-class ArithmeticAdditionGenericFixture : public framework::Fixture
+class ArithmeticOperationGenericFixture : public framework::Fixture
 {
 public:
     template <typename...>
-    void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy,
+    void setup(reference::ArithmeticOperation op, const TensorShape &shape0, const TensorShape &shape1,
+               DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy,
                QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
     {
+        _op        = op;
         _target    = compute_target(shape0, shape1, data_type0, data_type1, output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out);
         _reference = compute_reference(shape0, shape1, data_type0, data_type1, output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out);
     }
@@ -68,8 +70,8 @@ protected:
         TensorType dst      = create_tensor<TensorType>(TensorShape::broadcast_shape(shape0, shape1), output_data_type, 1, qinfo_out);
 
         // Create and configure function
-        FunctionType add;
-        add.configure(&ref_src1, &ref_src2, &dst, convert_policy);
+        FunctionType arith_op;
+        arith_op.configure(&ref_src1, &ref_src2, &dst, convert_policy);
 
         ARM_COMPUTE_EXPECT(ref_src1.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(ref_src2.info()->is_resizable(), framework::LogLevel::ERRORS);
@@ -89,7 +91,7 @@ protected:
         fill(AccessorType(ref_src2), 1);
 
         // Compute function
-        add.run();
+        arith_op.run();
 
         return dst;
     }
@@ -107,39 +109,40 @@ protected:
         fill(ref_src1, 0);
         fill(ref_src2, 1);
 
-        return reference::arithmetic_addition<T>(ref_src1, ref_src2, ref_dst, convert_policy);
+        return reference::arithmetic_operation<T>(_op, ref_src1, ref_src2, ref_dst, convert_policy);
     }
 
-    TensorType      _target{};
-    SimpleTensor<T> _reference{};
+    TensorType                     _target{};
+    SimpleTensor<T>                _reference{};
+    reference::ArithmeticOperation _op{ reference::ArithmeticOperation::ADD };
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-class ArithmeticAdditionBroadcastValidationFixture : public ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>
+class ArithmeticAdditionBroadcastValidationFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
 {
 public:
     template <typename...>
     void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
     {
-        ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(shape0, shape1, data_type0, data_type1,
-                                                                                           output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape0, shape1, data_type0, data_type1,
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
     }
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-class ArithmeticAdditionValidationFixture : public ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>
+class ArithmeticAdditionValidationFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
 {
 public:
     template <typename...>
     void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
     {
-        ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(shape, shape, data_type0, data_type1,
-                                                                                           output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape, shape, data_type0, data_type1,
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
     }
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-class ArithmeticAdditionValidationQuantizedFixture : public ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>
+class ArithmeticAdditionValidationQuantizedFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
 {
 public:
     template <typename...>
@@ -147,11 +150,52 @@ public:
                QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
 
     {
-        ArithmeticAdditionGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(shape, shape, data_type0, data_type1,
-                                                                                           output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out);
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape, shape, data_type0, data_type1,
+                                                                                            output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out);
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticSubtractionBroadcastValidationFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
+    {
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape0, shape1,
+                                                                                            data_type0, data_type1, output_data_type, convert_policy,
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticSubtractionValidationFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
+    {
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape, shape,
+                                                                                            data_type0, data_type1, output_data_type, convert_policy,
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticSubtractionValidationQuantizedFixture : public ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+    {
+        ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape, shape,
+                                                                                            data_type0, data_type1, output_data_type,
+                                                                                            convert_policy, qinfo0, qinfo1, qinfo_out);
     }
 };
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
-#endif /* ARM_COMPUTE_TEST_ARITHMETIC_ADDITION_FIXTURE */
+#endif /* ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_FIXTURE */
diff --git a/tests/validation/fixtures/ArithmeticSubtractionFixture.h b/tests/validation/fixtures/ArithmeticSubtractionFixture.h
deleted file mode 100644
index 9897677691..0000000000
--- a/tests/validation/fixtures/ArithmeticSubtractionFixture.h
+++ /dev/null
@@ -1,114 +0,0 @@
-/*
- * Copyright (c) 2017-2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_FIXTURE
-#define ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_FIXTURE
-
-#include "arm_compute/core/TensorShape.h"
-#include "arm_compute/core/Types.h"
-#include "tests/AssetsLibrary.h"
-#include "tests/Globals.h"
-#include "tests/IAccessor.h"
-#include "tests/framework/Asserts.h"
-#include "tests/framework/Fixture.h"
-#include "tests/validation/Helpers.h"
-#include "tests/validation/reference/ArithmeticSubtraction.h"
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-template <typename TensorType, typename AccessorType, typename FunctionType, typename T1, typename T2 = T1, typename T3 = T1>
-class ArithmeticSubtractionValidationFixture : public framework::Fixture
-{
-public:
-    template <typename...>
-    void setup(TensorShape shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
-    {
-        _target    = compute_target(shape, data_type0, data_type1, output_data_type, convert_policy);
-        _reference = compute_reference(shape, data_type0, data_type1, output_data_type, convert_policy);
-    }
-
-protected:
-    template <typename U>
-    void fill(U &&tensor, int i)
-    {
-        library->fill_tensor_uniform(tensor, i);
-    }
-
-    TensorType compute_target(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
-    {
-        // Create tensors
-        TensorType ref_src1 = create_tensor<TensorType>(shape, data_type0, 1);
-        TensorType ref_src2 = create_tensor<TensorType>(shape, data_type1, 1);
-        TensorType dst      = create_tensor<TensorType>(shape, output_data_type, 1);
-
-        // Create and configure function
-        FunctionType sub;
-        sub.configure(&ref_src1, &ref_src2, &dst, convert_policy);
-
-        ARM_COMPUTE_EXPECT(ref_src1.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(ref_src2.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(dst.info()->is_resizable(), framework::LogLevel::ERRORS);
-
-        // Allocate tensors
-        ref_src1.allocator()->allocate();
-        ref_src2.allocator()->allocate();
-        dst.allocator()->allocate();
-
-        ARM_COMPUTE_EXPECT(!ref_src1.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(!ref_src2.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
-
-        // Fill tensors
-        fill(AccessorType(ref_src1), 0);
-        fill(AccessorType(ref_src2), 1);
-
-        // Compute function
-        sub.run();
-
-        return dst;
-    }
-
-    SimpleTensor<T3> compute_reference(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
-    {
-        // Create reference
-        SimpleTensor<T1> ref_src1{ shape, data_type0, 1 };
-        SimpleTensor<T2> ref_src2{ shape, data_type1, 1 };
-
-        // Fill reference
-        fill(ref_src1, 0);
-        fill(ref_src2, 1);
-
-        return reference::arithmetic_subtraction<T1, T2, T3>(ref_src1, ref_src2, output_data_type, convert_policy);
-    }
-
-    TensorType       _target{};
-    SimpleTensor<T3> _reference{};
-};
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_FIXTURE */
diff --git a/tests/validation/fixtures/LSTMLayerFixture.h b/tests/validation/fixtures/LSTMLayerFixture.h
index 20df855242..bc892bfecf 100644
--- a/tests/validation/fixtures/LSTMLayerFixture.h
+++ b/tests/validation/fixtures/LSTMLayerFixture.h
@@ -28,8 +28,7 @@
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Fixture.h"
 #include "tests/validation/reference/ActivationLayer.h"
-#include "tests/validation/reference/ArithmeticAddition.h"
-#include "tests/validation/reference/ArithmeticSubtraction.h"
+#include "tests/validation/reference/ArithmeticOperations.h"
 #include "tests/validation/reference/FullyConnectedLayer.h"
 #include "tests/validation/reference/GEMM.h"
 #include "tests/validation/reference/PixelWiseMultiplication.h"
@@ -333,12 +332,12 @@ protected:
         SimpleTensor<T> fully_connected_forget = reference::fully_connected_layer(input, input_to_forget_w, forget_gate_bias, output_cell_shape);
         SimpleTensor<T> transposed_weights     = reference::transpose(recurrent_to_forget_w);
         SimpleTensor<T> gemm                   = reference::gemm(output_state_in, transposed_weights, cell_state_in, 1.f, 0.f);
-        SimpleTensor<T> forget_gate            = reference::arithmetic_addition(fully_connected_forget, gemm, data_type, ConvertPolicy::SATURATE);
+        SimpleTensor<T> forget_gate            = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, fully_connected_forget, gemm, data_type, ConvertPolicy::SATURATE);
 
         if(peephole_opt)
         {
             SimpleTensor<T> pixelwise_mul_forget_gate = reference::pixel_wise_multiplication(cell_state_in, cell_to_forget_w, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
-            forget_gate                               = reference::arithmetic_addition(forget_gate, pixelwise_mul_forget_gate, data_type, ConvertPolicy::SATURATE);
+            forget_gate                               = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, forget_gate, pixelwise_mul_forget_gate, data_type, ConvertPolicy::SATURATE);
         }
 
         forget_gate = reference::activation_layer(forget_gate, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
@@ -349,18 +348,18 @@ protected:
         {
             SimpleTensor<T> ones{ cell_bias_shape, data_type };
             fill_custom_val(ones, 1.f, 0);
-            input_gate = reference::arithmetic_subtraction<T, T, T>(ones, forget_gate, data_type, ConvertPolicy::SATURATE);
+            input_gate = reference::arithmetic_operation<T>(reference::ArithmeticOperation::SUB, ones, forget_gate, data_type, ConvertPolicy::SATURATE);
         }
         else
         {
             SimpleTensor<T> fully_connected_input = reference::fully_connected_layer(input, input_to_input_w, input_gate_bias, output_cell_shape);
             transposed_weights                    = reference::transpose(recurrent_to_input_w);
             gemm                                  = reference::gemm(output_state_in, transposed_weights, cell_state_in, 1.f, 0.f);
-            input_gate                            = reference::arithmetic_addition(fully_connected_input, gemm, data_type, ConvertPolicy::SATURATE);
+            input_gate                            = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, fully_connected_input, gemm, data_type, ConvertPolicy::SATURATE);
             if(peephole_opt)
             {
                 SimpleTensor<T> pixelwise_mul_input_gate = reference::pixel_wise_multiplication(cell_state_in, cell_to_input_w, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
-                input_gate                               = reference::arithmetic_addition(input_gate, pixelwise_mul_input_gate, data_type, ConvertPolicy::SATURATE);
+                input_gate                               = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, input_gate, pixelwise_mul_input_gate, data_type, ConvertPolicy::SATURATE);
             }
             input_gate = reference::activation_layer(input_gate, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
         }
@@ -370,10 +369,10 @@ protected:
         transposed_weights                         = reference::transpose(recurrent_to_cell_w);
         gemm                                       = reference::gemm(output_state_in, transposed_weights, cell_state_out, 1.f, 0.f);
         SimpleTensor<T> pixelwise_mul              = reference::pixel_wise_multiplication(cell_state_in, forget_gate, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
-        cell_state_out                             = reference::arithmetic_addition(fully_connected_cell_state, gemm, data_type, ConvertPolicy::SATURATE);
+        cell_state_out                             = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, fully_connected_cell_state, gemm, data_type, ConvertPolicy::SATURATE);
         cell_state_out                             = reference::activation_layer(cell_state_out, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
         cell_state_out                             = reference::pixel_wise_multiplication(cell_state_out, input_gate, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
-        cell_state_out                             = reference::arithmetic_addition(cell_state_out, pixelwise_mul, data_type, ConvertPolicy::SATURATE);
+        cell_state_out                             = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, cell_state_out, pixelwise_mul, data_type, ConvertPolicy::SATURATE);
         if(cell_threshold != 0.f)
         {
             cell_state_out = reference::activation_layer(cell_state_out, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LU_BOUNDED_RELU, -cell_threshold, cell_threshold));
@@ -383,11 +382,11 @@ protected:
         SimpleTensor<T> fully_connected_output = reference::fully_connected_layer(input, input_to_output_w, output_gate_bias, output_cell_shape);
         transposed_weights                     = reference::transpose(recurrent_to_output_w);
         gemm                                   = reference::gemm(output_state_in, transposed_weights, cell_state_out, 1.f, 0.f);
-        output                                 = reference::arithmetic_addition(fully_connected_output, gemm, data_type, ConvertPolicy::SATURATE);
+        output                                 = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, fully_connected_output, gemm, data_type, ConvertPolicy::SATURATE);
         if(peephole_opt)
         {
             pixelwise_mul = reference::pixel_wise_multiplication(cell_state_out, cell_to_output_w, 1, ConvertPolicy::SATURATE, RoundingPolicy::TO_NEAREST_EVEN);
-            output        = reference::arithmetic_addition(output, pixelwise_mul, data_type, ConvertPolicy::SATURATE);
+            output        = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, output, pixelwise_mul, data_type, ConvertPolicy::SATURATE);
         }
         output = reference::activation_layer(output, ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC));
 
diff --git a/tests/validation/fixtures/RNNLayerFixture.h b/tests/validation/fixtures/RNNLayerFixture.h
index 42b99cce1c..2645116b44 100644
--- a/tests/validation/fixtures/RNNLayerFixture.h
+++ b/tests/validation/fixtures/RNNLayerFixture.h
@@ -28,7 +28,7 @@
 #include "tests/framework/Asserts.h"
 #include "tests/framework/Fixture.h"
 #include "tests/validation/reference/ActivationLayer.h"
-#include "tests/validation/reference/ArithmeticAddition.h"
+#include "tests/validation/reference/ArithmeticOperations.h"
 #include "tests/validation/reference/FullyConnectedLayer.h"
 #include "tests/validation/reference/GEMM.h"
 
@@ -132,7 +132,7 @@ protected:
         SimpleTensor<T> out_w{ out_shape, data_type };
         SimpleTensor<T> fully_connected = reference::fully_connected_layer(input, weights, bias, out_shape);
         SimpleTensor<T> gemm            = reference::gemm(hidden_state, recurrent_weights, out_w, 1.f, 0.f);
-        SimpleTensor<T> add_res         = reference::arithmetic_addition(fully_connected, gemm, data_type, ConvertPolicy::SATURATE);
+        SimpleTensor<T> add_res         = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, fully_connected, gemm, data_type, ConvertPolicy::SATURATE);
         return reference::activation_layer(add_res, info);
     }
 
diff --git a/tests/validation/reference/ArithmeticAddition.cpp b/tests/validation/reference/ArithmeticOperations.cpp
index c68c6d4ed6..062be93b95 100644
--- a/tests/validation/reference/ArithmeticAddition.cpp
+++ b/tests/validation/reference/ArithmeticOperations.cpp
@@ -21,7 +21,7 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
-#include "ArithmeticAddition.h"
+#include "ArithmeticOperations.h"
 
 #include "arm_compute/core/Types.h"
 #include "tests/validation/Helpers.h"
@@ -37,11 +37,12 @@ namespace reference
 namespace
 {
 template <typename T>
-T add(T src1, T src2, ConvertPolicy convert_policy)
+T arithm_op(ArithmeticOperation op, T src1, T src2, ConvertPolicy convert_policy)
 {
     using intermediate_type = typename common_promoted_signed_type<T>::intermediate_type;
 
-    intermediate_type val = static_cast<intermediate_type>(src1) + static_cast<intermediate_type>(src2);
+    intermediate_type val = (op == ArithmeticOperation::ADD) ? static_cast<intermediate_type>(src1) + static_cast<intermediate_type>(src2) : static_cast<intermediate_type>
+                            (src1) - static_cast<intermediate_type>(src2);
 
     T result = (convert_policy == ConvertPolicy::SATURATE) ? saturate_cast<T>(val) : static_cast<T>(val);
 
@@ -52,7 +53,7 @@ template <size_t dim>
 struct BroadcastUnroll
 {
     template <typename T>
-    static void unroll(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst,
+    static void unroll(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst,
                        ConvertPolicy convert_policy, Coordinates &id_src1, Coordinates &id_src2, Coordinates &id_dst)
     {
         const bool src1_is_broadcast = (src1.shape()[dim - 1] != dst.shape()[dim - 1]);
@@ -64,7 +65,7 @@ struct BroadcastUnroll
 
         for(size_t i = 0; i < dst.shape()[dim - 1]; ++i, ++id_dst[dim - 1])
         {
-            BroadcastUnroll < dim - 1 >::unroll(src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
+            BroadcastUnroll < dim - 1 >::unroll(op, src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
 
             id_src1[dim - 1] += !src1_is_broadcast;
             id_src2[dim - 1] += !src2_is_broadcast;
@@ -76,26 +77,26 @@ template <>
 struct BroadcastUnroll<0>
 {
     template <typename T>
-    static void unroll(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst,
+    static void unroll(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst,
                        ConvertPolicy convert_policy, Coordinates &id_src1, Coordinates &id_src2, Coordinates &id_dst)
     {
-        dst[coord2index(dst.shape(), id_dst)] = add(src1[coord2index(src1.shape(), id_src1)], src2[coord2index(src2.shape(), id_src2)], convert_policy);
+        dst[coord2index(dst.shape(), id_dst)] = arithm_op(op, src1[coord2index(src1.shape(), id_src1)], src2[coord2index(src2.shape(), id_src2)], convert_policy);
     }
 };
 } // namespace
 
 template <typename T>
-SimpleTensor<T> arithmetic_addition(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst, ConvertPolicy convert_policy)
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst, ConvertPolicy convert_policy)
 {
     Coordinates id_src1, id_src2, id_dst;
 
-    BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
+    BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(op, src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
 
     return dst;
 }
 
 template <>
-SimpleTensor<uint8_t> arithmetic_addition(const SimpleTensor<uint8_t> &src1, const SimpleTensor<uint8_t> &src2, SimpleTensor<uint8_t> &dst, ConvertPolicy convert_policy)
+SimpleTensor<uint8_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<uint8_t> &src1, const SimpleTensor<uint8_t> &src2, SimpleTensor<uint8_t> &dst, ConvertPolicy convert_policy)
 {
     if(dst.data_type() == DataType::QASYMM8)
     {
@@ -105,7 +106,7 @@ SimpleTensor<uint8_t> arithmetic_addition(const SimpleTensor<uint8_t> &src1, con
 
         Coordinates id_src1, id_src2, id_dst;
 
-        BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(src1_tmp, src2_tmp, dst_tmp, convert_policy, id_src1, id_src2, id_dst);
+        BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(op, src1_tmp, src2_tmp, dst_tmp, convert_policy, id_src1, id_src2, id_dst);
 
         dst = convert_to_asymmetric(dst_tmp, dst.quantization_info());
         return dst;
@@ -115,31 +116,32 @@ SimpleTensor<uint8_t> arithmetic_addition(const SimpleTensor<uint8_t> &src1, con
         // DataType::U8
         Coordinates id_src1, id_src2, id_dst;
 
-        BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
+        BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(op, src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
 
         return dst;
     }
 }
 
-template SimpleTensor<int16_t> arithmetic_addition(const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, SimpleTensor<int16_t> &dst, ConvertPolicy convert_policy);
-template SimpleTensor<int8_t> arithmetic_addition(const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, SimpleTensor<int8_t> &dst, ConvertPolicy convert_policy);
-template SimpleTensor<half> arithmetic_addition(const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, SimpleTensor<half> &dst, ConvertPolicy convert_policy);
-template SimpleTensor<float> arithmetic_addition(const SimpleTensor<float> &src1, const SimpleTensor<float> &src2, SimpleTensor<float> &dst, ConvertPolicy convert_policy);
+template SimpleTensor<int16_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, SimpleTensor<int16_t> &dst,
+                                                    ConvertPolicy convert_policy);
+template SimpleTensor<int8_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, SimpleTensor<int8_t> &dst, ConvertPolicy convert_policy);
+template SimpleTensor<half> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, SimpleTensor<half> &dst, ConvertPolicy convert_policy);
+template SimpleTensor<float> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<float> &src1, const SimpleTensor<float> &src2, SimpleTensor<float> &dst, ConvertPolicy convert_policy);
 
 template <typename T>
-SimpleTensor<T> arithmetic_addition(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, DataType dst_data_type, ConvertPolicy convert_policy)
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, DataType dst_data_type, ConvertPolicy convert_policy)
 {
     ARM_COMPUTE_ERROR_ON_MSG(dst_data_type == DataType::QASYMM8, "For QASYMM8, the quantized output tensor should be passed directly.");
 
     SimpleTensor<T> dst(TensorShape::broadcast_shape(src1.shape(), src2.shape()), dst_data_type);
-    arithmetic_addition<T>(src1, src2, dst, convert_policy);
+    arithmetic_operation<T>(op, src1, src2, dst, convert_policy);
     return dst;
 }
 
-template SimpleTensor<int16_t> arithmetic_addition(const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int8_t> arithmetic_addition(const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<half> arithmetic_addition(const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<float> arithmetic_addition(const SimpleTensor<float> &src1, const SimpleTensor<float> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
+template SimpleTensor<int16_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
+template SimpleTensor<int8_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
+template SimpleTensor<half> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
+template SimpleTensor<float> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<float> &src1, const SimpleTensor<float> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
 
 } // namespace reference
 } // namespace validation
diff --git a/tests/validation/reference/ArithmeticAddition.h b/tests/validation/reference/ArithmeticOperations.h
index faeabd7a6f..73634823c4 100644
--- a/tests/validation/reference/ArithmeticAddition.h
+++ b/tests/validation/reference/ArithmeticOperations.h
@@ -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_ARITHMETIC_ADDITION_H__
-#define __ARM_COMPUTE_TEST_ARITHMETIC_ADDITION_H__
+#ifndef __ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_H__
+#define __ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_H__
 
 #include "tests/SimpleTensor.h"
 #include "tests/validation/Helpers.h"
@@ -35,14 +35,20 @@ namespace validation
 {
 namespace reference
 {
-template <typename T>
-SimpleTensor<T> arithmetic_addition(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst, ConvertPolicy convert_policy);
+/** Arithmetic operation types */
+enum class ArithmeticOperation
+{
+    ADD,
+    SUB
+};
 
 template <typename T>
-SimpleTensor<T> arithmetic_addition(const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst, ConvertPolicy convert_policy);
 
+template <typename T>
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
 } // namespace reference
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
-#endif /* __ARM_COMPUTE_TEST_ARITHMETIC_ADDITION_H__ */
+#endif /* __ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_H__ */
diff --git a/tests/validation/reference/ArithmeticSubtraction.cpp b/tests/validation/reference/ArithmeticSubtraction.cpp
deleted file mode 100644
index f39d01f9e8..0000000000
--- a/tests/validation/reference/ArithmeticSubtraction.cpp
+++ /dev/null
@@ -1,63 +0,0 @@
-/*
- * Copyright (c) 2017-2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "ArithmeticSubtraction.h"
-
-#include "tests/validation/Helpers.h"
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-namespace reference
-{
-template <typename T1, typename T2, typename T3>
-SimpleTensor<T3> arithmetic_subtraction(const SimpleTensor<T1> &src1, const SimpleTensor<T2> &src2, DataType dst_data_type, ConvertPolicy convert_policy)
-{
-    SimpleTensor<T3> result(src1.shape(), dst_data_type);
-
-    using intermediate_type = typename common_promoted_signed_type<typename std::conditional<sizeof(T1) >= sizeof(T2), T1, T2>::type >::intermediate_type;
-
-    for(int i = 0; i < src1.num_elements(); ++i)
-    {
-        intermediate_type val = static_cast<intermediate_type>(src1[i]) - static_cast<intermediate_type>(src2[i]);
-        result[i]             = (convert_policy == ConvertPolicy::SATURATE) ? saturate_cast<T3>(val) : static_cast<T3>(val);
-    }
-
-    return result;
-}
-
-template SimpleTensor<uint8_t> arithmetic_subtraction(const SimpleTensor<uint8_t> &src1, const SimpleTensor<uint8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int16_t> arithmetic_subtraction(const SimpleTensor<uint8_t> &src1, const SimpleTensor<uint8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int16_t> arithmetic_subtraction(const SimpleTensor<uint8_t> &src1, const SimpleTensor<int16_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int16_t> arithmetic_subtraction(const SimpleTensor<int16_t> &src1, const SimpleTensor<uint8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int16_t> arithmetic_subtraction(const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<int8_t> arithmetic_subtraction(const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<half> arithmetic_subtraction(const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-template SimpleTensor<float> arithmetic_subtraction(const SimpleTensor<float> &src1, const SimpleTensor<float> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-} // namespace reference
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
diff --git a/tests/validation/reference/ArithmeticSubtraction.h b/tests/validation/reference/ArithmeticSubtraction.h
deleted file mode 100644
index 9308314bda..0000000000
--- a/tests/validation/reference/ArithmeticSubtraction.h
+++ /dev/null
@@ -1,44 +0,0 @@
-/*
- * Copyright (c) 2017 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#ifndef __ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_H__
-#define __ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_H__
-
-#include "tests/SimpleTensor.h"
-#include "tests/validation/Helpers.h"
-
-namespace arm_compute
-{
-namespace test
-{
-namespace validation
-{
-namespace reference
-{
-template <typename T1, typename T2, typename T3>
-SimpleTensor<T3> arithmetic_subtraction(const SimpleTensor<T1> &src1, const SimpleTensor<T2> &src2, DataType dst_data_type, ConvertPolicy convert_policy);
-} // namespace reference
-} // namespace validation
-} // namespace test
-} // namespace arm_compute
-#endif /* __ARM_COMPUTE_TEST_ARITHMETIC_SUBTRACTION_H__ */
diff --git a/tests/validation/reference/LaplacianPyramid.cpp b/tests/validation/reference/LaplacianPyramid.cpp
index 566847453b..21ddc1e48e 100644
--- a/tests/validation/reference/LaplacianPyramid.cpp
+++ b/tests/validation/reference/LaplacianPyramid.cpp
@@ -23,7 +23,7 @@
  */
 #include "LaplacianPyramid.h"
 
-#include "tests/validation/reference/ArithmeticSubtraction.h"
+#include "tests/validation/reference/ArithmeticOperations.h"
 #include "tests/validation/reference/DepthConvertLayer.h"
 #include "tests/validation/reference/Gaussian5x5.h"
 #include "tests/validation/reference/GaussianPyramidHalf.h"
@@ -53,7 +53,10 @@ std::vector<SimpleTensor<U>> laplacian_pyramid(const SimpleTensor<T> &src, Simpl
         const SimpleTensor<T> level_filtered = reference::gaussian5x5(gaussian_level_pyramid[i], border_mode, constant_border_value);
         pyramid_conv.push_back(level_filtered);
 
-        const SimpleTensor<U> level_sub = reference::arithmetic_subtraction<T, T, U>(gaussian_level_pyramid[i], level_filtered, dst.data_type(), ConvertPolicy::WRAP);
+        const SimpleTensor<U> level_filtered_converted = depth_convert<T, U>(level_filtered, DataType::S16, ConvertPolicy::WRAP, 0);
+        const SimpleTensor<U> gaussian_level_converted = depth_convert<T, U>(gaussian_level_pyramid[i], DataType::S16, ConvertPolicy::WRAP, 0);
+
+        const SimpleTensor<U> level_sub = reference::arithmetic_operation<U>(reference::ArithmeticOperation::SUB, gaussian_level_converted, level_filtered_converted, dst.data_type(), ConvertPolicy::WRAP);
         pyramid_dst.push_back(level_sub);
     }
 
diff --git a/tests/validation/reference/LaplacianReconstruct.cpp b/tests/validation/reference/LaplacianReconstruct.cpp
index 23468287b0..ef14355bc8 100644
--- a/tests/validation/reference/LaplacianReconstruct.cpp
+++ b/tests/validation/reference/LaplacianReconstruct.cpp
@@ -24,7 +24,7 @@
 #include "LaplacianReconstruct.h"
 
 #include "arm_compute/core/Types.h"
-#include "tests/validation/reference/ArithmeticAddition.h"
+#include "tests/validation/reference/ArithmeticOperations.h"
 #include "tests/validation/reference/DepthConvertLayer.h"
 #include "tests/validation/reference/Scale.h"
 
@@ -45,7 +45,7 @@ SimpleTensor<U> laplacian_reconstruct(const std::vector<SimpleTensor<T>> &pyrami
     const DataType data_type  = low_res.data_type();
 
     // input + L(n-1)
-    tmp_pyramid[last_level] = reference::arithmetic_addition(low_res, pyramid[last_level], data_type, ConvertPolicy::SATURATE);
+    tmp_pyramid[last_level] = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, low_res, pyramid[last_level], data_type, ConvertPolicy::SATURATE);
 
     // Scale levels n-1 to 1, and add levels n-2 to 0
     for(size_t i = last_level; i-- > 0;)
@@ -56,7 +56,7 @@ SimpleTensor<U> laplacian_reconstruct(const std::vector<SimpleTensor<T>> &pyrami
         tmp_pyramid[i] = reference::scale(tmp_pyramid[i + 1], scale_x, scale_y, InterpolationPolicy::NEAREST_NEIGHBOR,
                                           border_mode, constant_border_value, SamplingPolicy::CENTER, false);
 
-        tmp_pyramid[i] = reference::arithmetic_addition(tmp_pyramid[i], pyramid[i], data_type, ConvertPolicy::SATURATE);
+        tmp_pyramid[i] = reference::arithmetic_operation(reference::ArithmeticOperation::ADD, tmp_pyramid[i], pyramid[i], data_type, ConvertPolicy::SATURATE);
     }
 
     return reference::depth_convert<T, U>(tmp_pyramid[0], DataType::U8, ConvertPolicy::SATURATE, 0);