COMPMID-2412: Add QSYMM16 support for ElementwiseAddition for CL

Arithmetic addition uses the same code as other element-wise operations.
Hence, adding QSYMM16 support for addition automatically adds the same
support for:

- arithmetic subtraction
- element-wise min
- element-wise max
- squared difference

Change-Id: If986102844f62e29dd23c03f9245910db43f9043
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1384
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Manuel Bottini <manuel.bottini@arm.com>
Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

6 files changed, 357 insertions, 140 deletions

diff --git a/tests/validation/CL/ArithmeticAddition.cpp b/tests/validation/CL/ArithmeticAddition.cpp
index bed04af9e5..e9b4b16b0c 100644
--- a/tests/validation/CL/ArithmeticAddition.cpp
+++ b/tests/validation/CL/ArithmeticAddition.cpp
@@ -50,6 +50,9 @@ const auto ArithmeticAdditionU8Dataset = combine(combine(framework::dataset::mak
 const auto ArithmeticAdditionQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
                                                       framework::dataset::make("DataType",
                                                                                DataType::QASYMM8));
+const auto ArithmeticAdditionQSYMM16Dataset = combine(combine(framework::dataset::make("DataType", DataType::QSYMM16), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                      framework::dataset::make("DataType",
+                                                                               DataType::QSYMM16));
 const auto ArithmeticAdditionS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
                                                   framework::dataset::make("DataType", DataType::S16));
 const auto ArithmeticAdditionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
@@ -93,6 +96,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLArithmeticAdditionFixture = ArithmeticAdditionValidationFixture<CLTensor, CLAccessor, CLArithmeticAddition, T>;
 
+TEST_SUITE(Integer)
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                shape, policy)
@@ -123,7 +127,48 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<uint8_t>, framework
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // U8
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+               shape, data_type, policy)
+{
+    // 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
+    CLArithmeticAddition 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(), 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, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // S16
+TEST_SUITE_END() // Integer
 
 template <typename T>
 using CLArithmeticAdditionQuantizedFixture = ArithmeticAdditionValidationQuantizedFixture<CLTensor, CLAccessor, CLArithmeticAddition, T>;
@@ -156,27 +201,22 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Sma
 FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ArithmeticAdditionQASYMM8Dataset),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
-                       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) }))
-
-                      )
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(1.f / 255.f, 5) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
-
-TEST_SUITE(S16)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
-               shape, data_type, policy)
+TEST_SUITE_END() // QASYMM8
+TEST_SUITE(QSYMM16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+               shape, policy)
 {
     // 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);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QSYMM16);
 
     // Create and Configure function
     CLArithmeticAddition add;
@@ -193,20 +233,18 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datas
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionS16Dataset),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
-
-FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticAdditionS16Dataset),
-                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
+                       ArithmeticAdditionQSYMM16Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // QSYMM16
+TEST_SUITE_END() // Quantized
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -216,7 +254,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<half>, framework::D
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // FP16
 
 TEST_SUITE(FP32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
@@ -274,11 +312,11 @@ FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLArithmeticAdditionBroadcastFixture<f
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // FP32
+TEST_SUITE_END() // Float
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // ArithmeticAddition
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/CL/ArithmeticSubtraction.cpp b/tests/validation/CL/ArithmeticSubtraction.cpp
index 796486bf02..b6a8139c0d 100644
--- a/tests/validation/CL/ArithmeticSubtraction.cpp
+++ b/tests/validation/CL/ArithmeticSubtraction.cpp
@@ -51,6 +51,9 @@ const auto ArithmeticSubtractionU8Dataset = combine(combine(framework::dataset::
 const auto ArithmeticSubtractionQASYMM8Dataset = combine(combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::QASYMM8)),
                                                          framework::dataset::make("DataType",
                                                                                   DataType::QASYMM8));
+const auto ArithmeticSubtractionQSYMM16Dataset = combine(combine(framework::dataset::make("DataType", DataType::QSYMM16), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                         framework::dataset::make("DataType",
+                                                                                  DataType::QSYMM16));
 const auto ArithmeticSubtractionS16Dataset = combine(combine(framework::dataset::make("DataType", { DataType::U8, DataType::S16 }), framework::dataset::make("DataType", DataType::S16)),
                                                      framework::dataset::make("DataType", DataType::S16));
 const auto ArithmeticSubtractionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
@@ -94,6 +97,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLArithmeticSubtractionFixture = ArithmeticSubtractionValidationFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
 
+TEST_SUITE(Integer)
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                shape, policy)
@@ -124,7 +128,48 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<uint8_t>, framew
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // U8
+
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
+                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+               shape, data_type, policy)
+{
+    // 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
+    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(), 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, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
+                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END() // S16
+TEST_SUITE_END() // Integer
 
 template <typename T>
 using CLArithmeticSubtractionQuantizedFixture = ArithmeticSubtractionValidationQuantizedFixture<CLTensor, CLAccessor, CLArithmeticSubtraction, T>;
@@ -157,31 +202,26 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Sma
 FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ArithmeticSubtractionQASYMM8Dataset),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
-                       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) }))
-
-                      )
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(1.f / 255.f, 5) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
-
-TEST_SUITE(S16)
-DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
-               shape, data_type, policy)
+TEST_SUITE_END() // QASYMM8
+TEST_SUITE(QSYMM16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+               shape, policy)
 {
     // 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);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QSYMM16);
 
     // Create and Configure function
-    CLArithmeticSubtraction add;
-    add.configure(&ref_src1, &ref_src2, &dst, policy);
+    CLArithmeticSubtraction sub;
+    sub.configure(&ref_src1, &ref_src2, &dst, policy);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -194,20 +234,18 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datas
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticSubtractionS16Dataset),
-                                                                                                                     framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
-{
-    // Validate output
-    validate(CLAccessor(_target), _reference);
-}
-
-FIXTURE_DATA_TEST_CASE(RunLarge, CLArithmeticSubtractionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticSubtractionS16Dataset),
-                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
+                       ArithmeticSubtractionQSYMM16Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // QSYMM16
+TEST_SUITE_END() // Quantized
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -217,7 +255,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticSubtractionFixture<half>, framework
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
+TEST_SUITE_END() // FP16
 
 TEST_SUITE(FP32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
@@ -275,11 +313,11 @@ FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, CLArithmeticSubtractionBroadcastFixtur
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // FP32
+TEST_SUITE_END() // Float
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // ArithmeticSubtraction
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseMax.cpp b/tests/validation/CL/ElementwiseMax.cpp
index 773a91905e..255acbe645 100644
--- a/tests/validation/CL/ElementwiseMax.cpp
+++ b/tests/validation/CL/ElementwiseMax.cpp
@@ -53,6 +53,9 @@ const auto ElementwiseMaxU8Dataset = combine(combine(framework::dataset::make("D
 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 ElementwiseMaxQSYMM16Dataset = combine(combine(framework::dataset::make("DataType", DataType::QSYMM16), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                  framework::dataset::make("DataType",
+                                                                           DataType::QSYMM16));
 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)),
@@ -96,6 +99,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLElementwiseMaxFixture = ElementwiseMaxValidationFixture<CLTensor, CLAccessor, CLElementwiseMax, T>;
 
+TEST_SUITE(Integer)
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
                shape)
@@ -106,8 +110,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
 
     // Create and Configure function
-    CLElementwiseMax add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMax max;
+    max.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -127,6 +131,38 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<uint8_t>, framework::Da
 }
 TEST_SUITE_END()
 
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), 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 max;
+    max.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::ALL, combine(datasets::SmallShapes(), ElementwiseMaxS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
 template <typename T>
 using CLElementwiseMaxQuantizedFixture = ElementwiseMaxValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseMax, T>;
 
@@ -141,8 +177,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8);
 
     // Create and Configure function
-    CLElementwiseMax add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMax max;
+    max.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -157,30 +193,26 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
 
 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) }))
-
-                      )
+                                                                                                                       framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                                                                                                                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                                                                                                                       framework::dataset::make("OutQInfo", { 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(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-               shape, data_type)
+TEST_SUITE(QSYMM16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
+               shape)
 {
     // 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);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QSYMM16);
 
     // Create and Configure function
-    CLElementwiseMax add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMax max;
+    max.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -193,12 +225,17 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Sma
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxFixture<int16_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseMaxS16Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMaxQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                                                                                                                       ElementwiseMaxQSYMM16Dataset),
+                                                                                                                       framework::dataset::make("Src0QInfo", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                                                                                                                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                                                                                                                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
 TEST_SUITE_END()
+TEST_SUITE_END()
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -219,8 +256,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
 
     // Create and Configure function
-    CLElementwiseMax add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMax max;
+    max.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -251,8 +288,8 @@ FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseMaxBroadcastFixture<float
 TEST_SUITE_END()
 TEST_SUITE_END()
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // ElementwiseMax
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseMin.cpp b/tests/validation/CL/ElementwiseMin.cpp
index b6486fccc7..01d18967f8 100644
--- a/tests/validation/CL/ElementwiseMin.cpp
+++ b/tests/validation/CL/ElementwiseMin.cpp
@@ -53,6 +53,9 @@ const auto ElementwiseMinU8Dataset = combine(combine(framework::dataset::make("D
 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 ElementwiseMinQSYMM16Dataset = combine(combine(framework::dataset::make("DataType", DataType::QSYMM16), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                  framework::dataset::make("DataType",
+                                                                           DataType::QSYMM16));
 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)),
@@ -96,6 +99,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLElementwiseMinFixture = ElementwiseMinValidationFixture<CLTensor, CLAccessor, CLElementwiseMin, T>;
 
+TEST_SUITE(Integer)
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
                shape)
@@ -106,8 +110,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
 
     // Create and Configure function
-    CLElementwiseMin add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMin min;
+    min.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -127,6 +131,38 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<uint8_t>, framework::Da
 }
 TEST_SUITE_END()
 
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), 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 min;
+    min.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::ALL, combine(datasets::SmallShapes(), ElementwiseMinS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
 template <typename T>
 using CLElementwiseMinQuantizedFixture = ElementwiseMinValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseMin, T>;
 
@@ -141,8 +177,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8);
 
     // Create and Configure function
-    CLElementwiseMin add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMin min;
+    min.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -157,30 +193,26 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
 
 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) }))
-
-                      )
+                                                                                                                       framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                                                                                                                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                                                                                                                       framework::dataset::make("OutQInfo", { 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(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-               shape, data_type)
+TEST_SUITE(QSYMM16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
+               shape)
 {
     // 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);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QSYMM16);
 
     // Create and Configure function
-    CLElementwiseMin add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMin min;
+    min.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -193,12 +225,17 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Sma
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinFixture<int16_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseMinS16Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseMinQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                                                                                                                       ElementwiseMinQSYMM16Dataset),
+                                                                                                                       framework::dataset::make("SrcQInfo0", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                                                                                                                       framework::dataset::make("SrcQInfo1", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                                                                                                                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
 }
 TEST_SUITE_END()
+TEST_SUITE_END()
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -219,8 +256,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
 
     // Create and Configure function
-    CLElementwiseMin add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseMin min;
+    min.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -250,8 +287,8 @@ FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseMinBroadcastFixture<float
 TEST_SUITE_END()
 TEST_SUITE_END()
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // ElementwiseMin
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/CL/ElementwiseSquaredDiff.cpp b/tests/validation/CL/ElementwiseSquaredDiff.cpp
index 35fecf127c..edc150109e 100644
--- a/tests/validation/CL/ElementwiseSquaredDiff.cpp
+++ b/tests/validation/CL/ElementwiseSquaredDiff.cpp
@@ -45,6 +45,7 @@ namespace
 {
 RelativeTolerance<float> tolerance_fp32(0.000001f);
 RelativeTolerance<float> tolerance_fp16(0.001f);
+AbsoluteTolerance<float> tolerance_qsymm16(1);
 
 constexpr unsigned int num_elems_processed_per_iteration = 16;
 /** Input data sets **/
@@ -54,6 +55,9 @@ const auto ElementwiseSquaredDiffU8Dataset = combine(combine(framework::dataset:
 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 ElementwiseSquaredDiffQSYMM16Dataset = combine(combine(framework::dataset::make("DataType", DataType::QSYMM16), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                          framework::dataset::make("DataType",
+                                                                                   DataType::QSYMM16));
 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)),
@@ -97,6 +101,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
 template <typename T>
 using CLElementwiseSquaredDiffFixture = ElementwiseSquaredDiffValidationFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
 
+TEST_SUITE(Integer)
 TEST_SUITE(U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
                shape)
@@ -107,8 +112,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
 
     // Create and Configure function
-    CLElementwiseSquaredDiff add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseSquaredDiff sqdiff;
+    sqdiff.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -128,6 +133,38 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<uint8_t>, frame
 }
 TEST_SUITE_END()
 
+TEST_SUITE(S16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), 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 sqdiff;
+    sqdiff.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::ALL, combine(datasets::SmallShapes(), ElementwiseSquaredDiffS16Dataset))
+{
+    // Validate output
+    validate(CLAccessor(_target), _reference);
+}
+TEST_SUITE_END()
+TEST_SUITE_END()
+
 template <typename T>
 using CLElementwiseSquaredDiffQuantizedFixture = ElementwiseSquaredDiffValidationQuantizedFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
 
@@ -142,8 +179,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8);
 
     // Create and Configure function
-    CLElementwiseSquaredDiff add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseSquaredDiff sqdiff;
+    sqdiff.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -158,30 +195,26 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
 
 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) }))
-
-                      )
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
+                       framework::dataset::make("OutQInfo", { 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(datasets::SmallShapes(), framework::dataset::make("DataType", { DataType::U8, DataType::S16 })),
-               shape, data_type)
+TEST_SUITE(QSYMM16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes(),
+               shape)
 {
     // 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);
+    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QSYMM16);
+    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QSYMM16);
 
     // Create and Configure function
-    CLElementwiseSquaredDiff add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseSquaredDiff sqdiff;
+    sqdiff.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -194,12 +227,17 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::Sma
     validate(dst.info()->padding(), padding);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<int16_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseSquaredDiffS16Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                       ElementwiseSquaredDiffQSYMM16Dataset),
+                       framework::dataset::make("Src0QInfo", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
 {
     // Validate output
-    validate(CLAccessor(_target), _reference);
+    validate(CLAccessor(_target), _reference, tolerance_qsymm16);
 }
 TEST_SUITE_END()
+TEST_SUITE_END()
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
@@ -220,8 +258,8 @@ DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, datasets::SmallShapes
     CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
 
     // Create and Configure function
-    CLElementwiseSquaredDiff add;
-    add.configure(&ref_src1, &ref_src2, &dst);
+    CLElementwiseSquaredDiff sqdiff;
+    sqdiff.configure(&ref_src1, &ref_src2, &dst);
 
     // Validate valid region
     const ValidRegion valid_region = shape_to_valid_region(shape);
@@ -251,8 +289,8 @@ FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseSquaredDiffBroadcastFixtu
 TEST_SUITE_END()
 TEST_SUITE_END()
 
-TEST_SUITE_END()
-TEST_SUITE_END()
+TEST_SUITE_END() // ElementwiseSquaredDiff
+TEST_SUITE_END() // CL
 } // namespace validation
 } // namespace test
 } // namespace arm_compute
diff --git a/tests/validation/reference/ElementwiseOperations.cpp b/tests/validation/reference/ElementwiseOperations.cpp
index 44eb417969..d5a37a0fae 100644
--- a/tests/validation/reference/ElementwiseOperations.cpp
+++ b/tests/validation/reference/ElementwiseOperations.cpp
@@ -184,10 +184,39 @@ SimpleTensor<uint8_t> arithmetic_operation(ArithmeticOperation op, const SimpleT
     }
 }
 
+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)
+{
+    if(dst.data_type() == DataType::QSYMM16)
+    {
+        SimpleTensor<float> src1_tmp = convert_from_symmetric<int16_t>(src1);
+        SimpleTensor<float> src2_tmp = convert_from_symmetric<int16_t>(src2);
+        SimpleTensor<float> dst_tmp(TensorShape::broadcast_shape(src1.shape(), src2.shape()), dst.data_type());
+
+        Coordinates id_src1{};
+        Coordinates id_src2{};
+        Coordinates 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_symmetric<int16_t>(dst_tmp, dst.quantization_info());
+        return dst;
+    }
+    else
+    {
+        // DataType::S16
+        Coordinates id_src1{};
+        Coordinates id_src2{};
+        Coordinates id_dst{};
+
+        BroadcastUnroll<Coordinates::num_max_dimensions>::unroll(op, src1, src2, dst, convert_policy, id_src1, id_src2, id_dst);
+
+        return dst;
+    }
+}
+
 template SimpleTensor<int32_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int32_t> &src1, const SimpleTensor<int32_t> &src2, SimpleTensor<int32_t> &dst,
                                                     ConvertPolicy convert_policy);
-template SimpleTensor<int16_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int16_t> &src1, const SimpleTensor<int16_t> &src2, SimpleTensor<int16_t> &dst,
-                                                    ConvertPolicy convert_policy);
 template SimpleTensor<int8_t> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<int8_t> &src1, const SimpleTensor<int8_t> &src2, SimpleTensor<int8_t> &dst,
                                                    ConvertPolicy convert_policy);
 template SimpleTensor<half> arithmetic_operation(ArithmeticOperation op, const SimpleTensor<half> &src1, const SimpleTensor<half> &src2, SimpleTensor<half> &dst, ConvertPolicy convert_policy);
@@ -196,7 +225,7 @@ template SimpleTensor<float> arithmetic_operation(ArithmeticOperation op, const
 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.");
+    ARM_COMPUTE_ERROR_ON_MSG(is_data_type_quantized(dst_data_type), "For quantized data types, 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);