COMPMID-1717: CL: Implement Maximum, Minimum, SquaredDifference

Change-Id: Ice653e48211053bd3cd20a693bd76de6b4efc370
Reviewed-on: https://review.mlplatform.org/270
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

9 files changed, 1543 insertions, 65 deletions

diff --git a/tests/validation/CL/ArithmeticAddition.cpp b/tests/validation/CL/ArithmeticAddition.cpp
index 09f1b7c5a9..6f7aa94521 100644
--- a/tests/validation/CL/ArithmeticAddition.cpp
+++ b/tests/validation/CL/ArithmeticAddition.cpp
@@ -24,7 +24,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "arm_compute/runtime/CL/functions/CLArithmeticAddition.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/PaddingCalculator.h"
 #include "tests/datasets/ConvertPolicyDataset.h"
@@ -43,7 +43,7 @@ namespace validation
 {
 namespace
 {
-constexpr unsigned int num_elems_processed_per_iteration = 8;
+constexpr unsigned int num_elems_processed_per_iteration = 16;
 /** Input data sets **/
 const auto ArithmeticAdditionU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)), framework::dataset::make("DataType",
                                                  DataType::U8));
diff --git a/tests/validation/CL/ArithmeticDivision.cpp b/tests/validation/CL/ArithmeticDivision.cpp
index 5d4fa1fd5e..87039d775f 100644
--- a/tests/validation/CL/ArithmeticDivision.cpp
+++ b/tests/validation/CL/ArithmeticDivision.cpp
@@ -24,7 +24,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "arm_compute/runtime/CL/functions/CLArithmeticDivision.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/PaddingCalculator.h"
 #include "tests/datasets/ConvertPolicyDataset.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/ArithmeticDivisionFixture.h"
+#include "tests/validation/fixtures/ElementwiseOperationsFixture.h"
 
 namespace arm_compute
 {
@@ -45,6 +45,20 @@ namespace
 {
 RelativeTolerance<float> tolerance_fp32(0.000001f);
 RelativeTolerance<float> tolerance_fp16(0.001f);
+
+constexpr unsigned int num_elems_processed_per_iteration = 16;
+/** Input data sets **/
+const auto ArithmeticDivisionU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)), framework::dataset::make("DataType",
+                                                 DataType::U8));
+const auto ArithmeticDivisionQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                      framework::dataset::make("DataType",
+                                                                               DataType::QASYMM8));
+const auto ArithmeticDivisionS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
+                                                  framework::dataset::make("DataType", DataType::S16));
+const auto ArithmeticDivisionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
+                                                   framework::dataset::make("DataType", DataType::F16));
+const auto ArithmeticDivisionFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
+                                                   framework::dataset::make("DataType", DataType::F32));
 } // namespace
 
 TEST_SUITE(CL)
@@ -53,25 +67,25 @@ TEST_SUITE(ArithmeticDivision)
 // *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),      // Wrong data type
+               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
-                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
                                                       }),
                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),
-                                                       TensorInfo(TensorShape(32U, 13U, 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),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
                                                      })),
-               framework::dataset::make("Expected", { false, false, false, false, true })),
+               framework::dataset::make("Expected", { true, true, false, false, false})),
                input1_info, input2_info, output_info, expected)
 {
     ARM_COMPUTE_EXPECT(bool(CLArithmeticDivision::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false))) == expected, framework::LogLevel::ERRORS);
@@ -82,17 +96,128 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLArithmeticDivisionFixture = ArithmeticDivisionValidationFixture<CLTensor, CLAccessor, CLArithmeticDivision, T>;
 
+TEST_SUITE(U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
+
+    // Create and Configure function
+    CLArithmeticDivision add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ArithmeticDivisionU8Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+template <typename T>
+using CLArithmeticDivisionQuantizedFixture = ArithmeticDivisionValidationQuantizedFixture<CLTensor, CLAccessor, CLArithmeticDivision, T>;
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    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
+    CLArithmeticDivision add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                       ArithmeticDivisionQASYMM8Dataset),
+                       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) }))
+
+                      )
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32, 0.01);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+               shape, data_type)
+{
+    // 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);
+
+    // Create and Configure function
+    CLArithmeticDivision add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ArithmeticDivisionS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticDivisionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ArithmeticDivisionS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::F16)))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ArithmeticDivisionFP16Dataset))
 {
     // Validate output
-    validate(CLAccessor(_target), _reference, tolerance_fp16);
+    validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
 }
-TEST_SUITE_END() // FP16
+TEST_SUITE_END()
 
 TEST_SUITE(FP32)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, concat(datasets::SmallShapes(), datasets::LargeShapes()), shape)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
 {
     // Create tensors
     CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::F32);
@@ -100,27 +225,27 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, concat(datasets::Smal
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
 
     // Create and Configure function
-    CLArithmeticDivision div;
-    div.configure(&ref_src1, &ref_src2, &dst);
+    CLArithmeticDivision add;
+    add.configure(&ref_src1, &ref_src2, &dst);
 
     // 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();
+    const PaddingSize padding = PaddingCalculator(shape.x(), num_elems_processed_per_iteration).required_padding();
     validate(ref_src1.info()->padding(), padding);
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::F32)))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticDivisionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ArithmeticDivisionFP32Dataset))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticDivisionFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), framework::dataset::make("DataType", DataType::F32)))
+FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticDivisionFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ArithmeticDivisionFP32Dataset))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
@@ -130,23 +255,23 @@ template <typename T>
 using CLArithmeticDivisionBroadcastFixture = ArithmeticDivisionBroadcastValidationFixture<CLTensor, CLAccessor, CLArithmeticDivision, T>;
 
 FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLArithmeticDivisionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapesBroadcast(),
-                       framework::dataset::make("DataType", DataType::F32)))
+                       ArithmeticDivisionFP32Dataset))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
 }
 
 FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLArithmeticDivisionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapesBroadcast(),
-                       framework::dataset::make("DataType", DataType::F32)))
+                       ArithmeticDivisionFP32Dataset))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
 }
-TEST_SUITE_END() // FP32
-TEST_SUITE_END() // Float
+TEST_SUITE_END()
+TEST_SUITE_END()
 
-TEST_SUITE_END() // ArithmeticDivision
-TEST_SUITE_END() // CL
+TEST_SUITE_END()
+TEST_SUITE_END()
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/CL/ArithmeticSubtraction.cpp b/tests/validation/CL/ArithmeticSubtraction.cpp
index cd13f42ec4..2cf410f373 100644
--- a/tests/validation/CL/ArithmeticSubtraction.cpp
+++ b/tests/validation/CL/ArithmeticSubtraction.cpp
@@ -24,7 +24,7 @@
 #include "arm_compute/core/Types.h"
 #include "arm_compute/runtime/CL/CLTensor.h"
 #include "arm_compute/runtime/CL/CLTensorAllocator.h"
-#include "arm_compute/runtime/CL/functions/CLArithmeticSubtraction.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
 #include "tests/CL/CLAccessor.h"
 #include "tests/PaddingCalculator.h"
 #include "tests/datasets/ConvertPolicyDataset.h"
@@ -43,6 +43,7 @@ namespace validation
 {
 namespace
 {
+constexpr unsigned int num_elems_processed_per_iteration = 16;
 /** Input data sets **/
 const auto ArithmeticSubtractionU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)),
                                                     framework::dataset::make("DataType",
@@ -64,26 +65,26 @@ 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})),
-        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);
 }
@@ -103,15 +104,15 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::da
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
 
     // 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);
     validate(dst.info()->valid_region(), valid_region);
 
     // Validate padding
-    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    const PaddingSize padding = PaddingCalculator(shape.x(), num_elems_processed_per_iteration).required_padding();
     validate(ref_src1.info()->padding(), padding);
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
@@ -123,7 +124,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<uint8_t>, framew
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END() // U8
+TEST_SUITE_END()
 
 template <typename T>
 using CLArithmeticSubtractionQuantizedFixture = ArithmeticSubtractionValidationQuantizedFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
@@ -147,7 +148,7 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::da
     validate(dst.info()->valid_region(), valid_region);
 
     // Validate padding
-    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    const PaddingSize padding = PaddingCalculator(shape.x(), num_elems_processed_per_iteration).required_padding();
     validate(ref_src1.info()->padding(), padding);
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
@@ -165,8 +166,8 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionQuantizedFixture<uint8_t
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END() // QASYMM8
-TEST_SUITE_END() // Quantized
+TEST_SUITE_END()
+TEST_SUITE_END()
 
 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 })),
@@ -179,15 +180,15 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(frame
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::S16);
 
     // 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);
     validate(dst.info()->valid_region(), valid_region);
 
     // Validate padding
-    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    const PaddingSize padding = PaddingCalculator(shape.x(), num_elems_processed_per_iteration).required_padding();
     validate(ref_src1.info()->padding(), padding);
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
@@ -206,7 +207,7 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionFixture<int16_t>, framew
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END() // S16
+TEST_SUITE_END()
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -216,7 +217,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<half>, framework
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END() // FP16
+TEST_SUITE_END()
 
 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 })),
@@ -228,15 +229,15 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::da
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
 
     // 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);
     validate(dst.info()->valid_region(), valid_region);
 
     // Validate padding
-    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    const PaddingSize padding = PaddingCalculator(shape.x(), num_elems_processed_per_iteration).required_padding();
     validate(ref_src1.info()->padding(), padding);
     validate(ref_src2.info()->padding(), padding);
     validate(dst.info()->padding(), padding);
@@ -274,11 +275,11 @@ FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLArithmeticSubtractionBroadcastFixtur
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END() // FP32
-TEST_SUITE_END() // Float
+TEST_SUITE_END()
+TEST_SUITE_END()
 
-TEST_SUITE_END() // ArithmeticSubtraction
-TEST_SUITE_END() // CL
+TEST_SUITE_END()
+TEST_SUITE_END()
 } // namespace validation
 } // namespace test
-} // namespace arm_compute
\ No newline at end of file
+} // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseMax.cpp b/tests/validation/CL/ElementwiseMax.cpp
new file mode 100644
index 0000000000..894688fe2c
--- /dev/null
+++ b/tests/validation/CL/ElementwiseMax.cpp
@@ -0,0 +1,277 @@
+/*
+ * Copyright (c) 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 "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/ConvertPolicyDataset.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ElementwiseOperationsFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+RelativeTolerance<float> tolerance_fp32(0.000001f);
+RelativeTolerance<float> tolerance_fp16(0.001f);
+
+constexpr unsigned int num_elems_processed_per_iteration = 16;
+/** Input data sets **/
+const auto ElementwiseMaxU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)), framework::dataset::make("DataType",
+                                             DataType::U8));
+const auto ElementwiseMaxQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                  framework::dataset::make("DataType",
+                                                                           DataType::QASYMM8));
+const auto ElementwiseMaxS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
+                                              framework::dataset::make("DataType", DataType::S16));
+const auto ElementwiseMaxFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
+                                               framework::dataset::make("DataType", DataType::F16));
+const auto ElementwiseMaxFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
+                                               framework::dataset::make("DataType", DataType::F32));
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(ElementwiseMax)
+
+// *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)
+{
+    ARM_COMPUTE_EXPECT(bool(CLElementwiseMax::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false))) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLElementwiseMaxFixture = ElementwiseMaxValidationFixture<CLTensor, CLAccessor, CLElementwiseMax, T>;
+
+TEST_SUITE(U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
+
+    // Create and Configure function
+    CLElementwiseMax add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMaxU8Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+template <typename T>
+using CLElementwiseMaxQuantizedFixture = ElementwiseMaxValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseMax, T>;
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    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
+    CLElementwiseMax add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                                                                                                                       ElementwiseMaxQASYMM8Dataset),
+                                                                                                                       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) }))
+
+                      )
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32, 0.01);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+               shape, data_type)
+{
+    // 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);
+
+    // Create and Configure function
+    CLElementwiseMax add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMaxS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseMaxFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseMaxS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(Float)
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseMaxFP16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(FP32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
+
+    // Create and Configure function
+    CLElementwiseMax add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMaxFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseMaxFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseMaxFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+template <typename T>
+using CLElementwiseMaxBroadcastFixture = ElementwiseMaxBroadcastValidationFixture<CLTensor, CLAccessor, CLElementwiseMax, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseMaxBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapesBroadcast(),
+                       ElementwiseMaxFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLElementwiseMaxBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapesBroadcast(),
+                       ElementwiseMaxFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseMin.cpp b/tests/validation/CL/ElementwiseMin.cpp
new file mode 100644
index 0000000000..05abfc853f
--- /dev/null
+++ b/tests/validation/CL/ElementwiseMin.cpp
@@ -0,0 +1,277 @@
+/*
+ * Copyright (c) 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 "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/ConvertPolicyDataset.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ElementwiseOperationsFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+RelativeTolerance<float> tolerance_fp32(0.000001f);
+RelativeTolerance<float> tolerance_fp16(0.001f);
+
+constexpr unsigned int num_elems_processed_per_iteration = 16;
+/** Input data sets **/
+const auto ElementwiseMinU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)), framework::dataset::make("DataType",
+                                             DataType::U8));
+const auto ElementwiseMinQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                  framework::dataset::make("DataType",
+                                                                           DataType::QASYMM8));
+const auto ElementwiseMinS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
+                                              framework::dataset::make("DataType", DataType::S16));
+const auto ElementwiseMinFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
+                                               framework::dataset::make("DataType", DataType::F16));
+const auto ElementwiseMinFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
+                                               framework::dataset::make("DataType", DataType::F32));
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(ElementwiseMin)
+
+// *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)
+{
+    ARM_COMPUTE_EXPECT(bool(CLElementwiseMin::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false))) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLElementwiseMinFixture = ElementwiseMinValidationFixture<CLTensor, CLAccessor, CLElementwiseMin, T>;
+
+TEST_SUITE(U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
+
+    // Create and Configure function
+    CLElementwiseMin add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMinU8Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+template <typename T>
+using CLElementwiseMinQuantizedFixture = ElementwiseMinValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseMin, T>;
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    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
+    CLElementwiseMin add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                                                                                                                       ElementwiseMinQASYMM8Dataset),
+                                                                                                                       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) }))
+
+                      )
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32, 0.01);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+               shape, data_type)
+{
+    // 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);
+
+    // Create and Configure function
+    CLElementwiseMin add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMinS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseMinFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseMinS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(Float)
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseMinFP16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(FP32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
+
+    // Create and Configure function
+    CLElementwiseMin add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseMinFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseMinFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseMinFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+template <typename T>
+using CLElementwiseMinBroadcastFixture = ElementwiseMinBroadcastValidationFixture<CLTensor, CLAccessor, CLElementwiseMin, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseMinBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapesBroadcast(),
+                       ElementwiseMinFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLElementwiseMinBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapesBroadcast(),
+                       ElementwiseMinFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseSquaredDiff.cpp b/tests/validation/CL/ElementwiseSquaredDiff.cpp
new file mode 100644
index 0000000000..c00f95b885
--- /dev/null
+++ b/tests/validation/CL/ElementwiseSquaredDiff.cpp
@@ -0,0 +1,278 @@
+/*
+ * Copyright (c) 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 "arm_compute/core/Types.h"
+#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/CLTensorAllocator.h"
+#include "arm_compute/runtime/CL/functions/CLElementwiseOperations.h"
+#include "tests/CL/CLAccessor.h"
+#include "tests/PaddingCalculator.h"
+#include "tests/datasets/ConvertPolicyDataset.h"
+#include "tests/datasets/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ElementwiseOperationsFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+RelativeTolerance<float> tolerance_fp32(0.000001f);
+RelativeTolerance<float> tolerance_fp16(0.001f);
+
+constexpr unsigned int num_elems_processed_per_iteration = 16;
+/** Input data sets **/
+const auto ElementwiseSquaredDiffU8Dataset = combine(combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U8)),
+                                                     framework::dataset::make("DataType",
+                                                                              DataType::U8));
+const auto ElementwiseSquaredDiffQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                          framework::dataset::make("DataType",
+                                                                                   DataType::QASYMM8));
+const auto ElementwiseSquaredDiffS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
+                                                      framework::dataset::make("DataType", DataType::S16));
+const auto ElementwiseSquaredDiffFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
+                                                       framework::dataset::make("DataType", DataType::F16));
+const auto ElementwiseSquaredDiffFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
+                                                       framework::dataset::make("DataType", DataType::F32));
+} // namespace
+
+TEST_SUITE(CL)
+TEST_SUITE(ElementwiseSquaredDiff)
+
+// *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)
+{
+    ARM_COMPUTE_EXPECT(bool(CLElementwiseSquaredDiff::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false))) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+template <typename T>
+using CLElementwiseSquaredDiffFixture = ElementwiseSquaredDiffValidationFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
+
+TEST_SUITE(U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::U8);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
+
+    // Create and Configure function
+    CLElementwiseSquaredDiff add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseSquaredDiffU8Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+template <typename T>
+using CLElementwiseSquaredDiffQuantizedFixture = ElementwiseSquaredDiffValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
+
+TEST_SUITE(Quantized)
+TEST_SUITE(QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    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
+    CLElementwiseSquaredDiff add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                       ElementwiseSquaredDiffQASYMM8Dataset),
+                       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) }))
+
+                      )
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32, 0.01);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+               shape, data_type)
+{
+    // 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);
+
+    // Create and Configure function
+    CLElementwiseSquaredDiff add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseSquaredDiffS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseSquaredDiffFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseSquaredDiffS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(Float)
+TEST_SUITE(FP16)
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
+}
+TEST_SUITE_END()
+
+TEST_SUITE(FP32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+               shape)
+{
+    // Create tensors
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::F32);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
+
+    // Create and Configure function
+    CLElementwiseSquaredDiff add;
+    add.configure(&ref_src1, &ref_src2, &dst);
+
+    // 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(), num_elems_processed_per_iteration).required_padding();
+    validate(ref_src1.info()->padding(), padding);
+    validate(ref_src2.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLElementwiseSquaredDiffFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseSquaredDiffFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+template <typename T>
+using CLElementwiseSquaredDiffBroadcastFixture = ElementwiseSquaredDiffBroadcastValidationFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseSquaredDiffBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapesBroadcast(),
+                       ElementwiseSquaredDiffFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLElementwiseSquaredDiffBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapesBroadcast(),
+                       ElementwiseSquaredDiffFP32Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference, tolerance_fp32);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
+TEST_SUITE_END()
+TEST_SUITE_END()
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/ElementwiseOperationsFixture.h b/tests/validation/fixtures/ElementwiseOperationsFixture.h
new file mode 100644
index 0000000000..b051c858c2
--- /dev/null
+++ b/tests/validation/fixtures/ElementwiseOperationsFixture.h
@@ -0,0 +1,286 @@
+/*
+ * Copyright (c) 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_ELEMENTWISE_OPERATIONS_FIXTURE
+#define ARM_COMPUTE_TEST_ELEMENTWISE_OPERATIONS_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/ElementwiseOperations.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticOperationsGenericFixture : public framework::Fixture
+{
+public:
+    template <typename...>
+    void setup(ArithmeticOperation op, const TensorShape &shape0, const TensorShape &shape1,
+               DataType data_type0, DataType data_type1, DataType output_data_type,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+    {
+        _op        = op;
+        _target    = compute_target(shape0, shape1, data_type0, data_type1, output_data_type, qinfo0, qinfo1, qinfo_out);
+        _reference = compute_reference(shape0, shape1, data_type0, data_type1, output_data_type, qinfo0, qinfo1, qinfo_out);
+    }
+
+protected:
+    template <typename U>
+    void fill(U &&tensor, int i)
+    {
+        library->fill_tensor_uniform(tensor, i);
+    }
+
+    TensorType compute_target(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type,
+                              QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+    {
+        // Create tensors
+        TensorType ref_src1 = create_tensor<TensorType>(shape0, data_type0, 1, qinfo0);
+        TensorType ref_src2 = create_tensor<TensorType>(shape1, data_type1, 1, qinfo1);
+        TensorType dst      = create_tensor<TensorType>(TensorShape::broadcast_shape(shape0, shape1), output_data_type, 1, qinfo_out);
+
+        // Create and configure function
+        FunctionType elem_op;
+        elem_op.configure(&ref_src1, &ref_src2, &dst);
+
+        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
+        elem_op.run();
+
+        return dst;
+    }
+
+    SimpleTensor<T> compute_reference(const TensorShape &shape0, const TensorShape &shape1,
+                                      DataType data_type0, DataType data_type1, DataType output_data_type,
+                                      QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+    {
+        // Create reference
+        SimpleTensor<T> ref_src1{ shape0, data_type0, 1, qinfo0 };
+        SimpleTensor<T> ref_src2{ shape1, data_type1, 1, qinfo1 };
+        SimpleTensor<T> ref_dst{ TensorShape::broadcast_shape(shape0, shape1), output_data_type, 1, qinfo_out };
+
+        // Fill reference
+        fill(ref_src1, 0);
+        fill(ref_src2, 1);
+
+        return reference::arithmetic_operation<T>(_op, ref_src1, ref_src2, ref_dst);
+    }
+
+    TensorType          _target{};
+    SimpleTensor<T>     _reference{};
+    ArithmeticOperation _op{ ArithmeticOperation::ADD };
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticDivisionBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<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)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape0, shape1,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticDivisionValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ArithmeticDivisionValidationQuantizedFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             qinfo0, qinfo1, qinfo_out);
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMaxBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<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)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MAX, shape0, shape1,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMaxValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MAX, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMaxValidationQuantizedFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MAX, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             qinfo0, qinfo1, qinfo_out);
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMinBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<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)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MIN, shape0, shape1,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMinValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MIN, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseMinValidationQuantizedFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::MIN, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             qinfo0, qinfo1, qinfo_out);
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseSquaredDiffBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<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)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::SQUARED_DIFF, shape0, shape1,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseSquaredDiffValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type)
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::SQUARED_DIFF, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+    }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseSquaredDiffValidationQuantizedFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+    template <typename...>
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type,
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+    {
+        ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::SQUARED_DIFF, shape, shape,
+                                                                                             data_type0, data_type1, output_data_type,
+                                                                                             qinfo0, qinfo1, qinfo_out);
+    }
+};
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* ARM_COMPUTE_TEST_ARITHMETIC_OPERATIONS_FIXTURE */
diff --git a/tests/validation/reference/ElementwiseOperations.cpp b/tests/validation/reference/ElementwiseOperations.cpp
new file mode 100644
index 0000000000..fe0467fe5e
--- /dev/null
+++ b/tests/validation/reference/ElementwiseOperations.cpp
@@ -0,0 +1,187 @@
+/*
+ * Copyright (c) 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 "ElementwiseOperations.h"
+
+#include "arm_compute/core/Types.h"
+#include "tests/validation/Helpers.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+namespace
+{
+template <typename T>
+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;
+
+    if(op == ArithmeticOperation::ADD)
+    {
+        val = static_cast<intermediate_type>(src1) + static_cast<intermediate_type>(src2);
+    }
+    else if(op == ArithmeticOperation::SUB)
+    {
+        val = static_cast<intermediate_type>(src1) - static_cast<intermediate_type>(src2);
+    }
+    else if(op == ArithmeticOperation::MIN)
+    {
+        val = std::min(static_cast<intermediate_type>(src1), static_cast<intermediate_type>(src2));
+    }
+    else if(op == ArithmeticOperation::MAX)
+    {
+        val = std::max(static_cast<intermediate_type>(src1), static_cast<intermediate_type>(src2));
+    }
+    else if(op == ArithmeticOperation::SQUARED_DIFF)
+    {
+        intermediate_type tmp = (static_cast<intermediate_type>(src1) - static_cast<intermediate_type>(src2));
+        val                   = tmp * tmp;
+    }
+    else if(op == ArithmeticOperation::DIV)
+    {
+        val = (static_cast<intermediate_type>(src1) / static_cast<intermediate_type>(src2));
+    }
+    else
+    {
+        ARM_COMPUTE_ERROR("Not handled");
+    }
+
+    T result;
+    if(op == ArithmeticOperation::ADD || op == ArithmeticOperation::SUB)
+    {
+        result = (convert_policy == ConvertPolicy::SATURATE) ? saturate_cast<T>(val) : static_cast<T>(val);
+    }
+    else
+    {
+        result = static_cast<T>(val);
+    }
+    return result;
+}
+
+template <size_t dim>
+struct BroadcastUnroll
+{
+    template <typename T>
+    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]);
+        const bool src2_is_broadcast = (src2.shape()[dim - 1] != dst.shape()[dim - 1]);
+
+        id_src1.set(dim - 1, 0);
+        id_src2.set(dim - 1, 0);
+        id_dst.set(dim - 1, 0);
+
+        for(size_t i = 0; i < dst.shape()[dim - 1]; ++i, ++id_dst[dim - 1])
+        {
+            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;
+        }
+    }
+};
+
+template <>
+struct BroadcastUnroll<0>
+{
+    template <typename T>
+    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)] = arithm_op(op, src1[coord2index(src1.shape(), id_src1)], src2[coord2index(src2.shape(), id_src2)], convert_policy);
+    }
+};
+} // namespace
+
+template <typename T>
+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(op, src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
+
+    return dst;
+}
+
+template <>
+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)
+    {
+        SimpleTensor<float> src1_tmp = convert_from_asymmetric(src1);
+        SimpleTensor<float> src2_tmp = convert_from_asymmetric(src2);
+        SimpleTensor<float> dst_tmp(TensorShape::broadcast_shape(src1.shape(), src2.shape()), dst.data_type());
+
+        Coordinates 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;
+    }
+    else
+    {
+        // DataType::U8
+        Coordinates 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_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_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_operation<T>(op, src1, src2, dst, convert_policy);
+    return dst;
+}
+
+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
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/reference/ElementwiseOperations.h b/tests/validation/reference/ElementwiseOperations.h
new file mode 100644
index 0000000000..7518ec86d5
--- /dev/null
+++ b/tests/validation/reference/ElementwiseOperations.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 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_ELEMENTWISE_OPERATIONS_H__
+#define __ARM_COMPUTE_TEST_ELEMENTWISE_OPERATIONS_H__
+
+#include "tests/SimpleTensor.h"
+#include "tests/validation/Helpers.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace reference
+{
+template <typename T>
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, SimpleTensor<T> &dst, ConvertPolicy convert_policy = ConvertPolicy::WRAP);
+
+template <typename T>
+SimpleTensor<T> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<T> &src1, const SimpleTensor<T> &src2, DataType dst_data_type, ConvertPolicy convert_policy = ConvertPolicy::WRAP);
+} // namespace reference
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_TEST_ELEMENTWISE_OPERATIONS_H__ */