COMPMID-2252 NECast.

Change-Id: I7532aea6827a325eb8457132d4787ac527e93cd4
Signed-off-by: Usama Arif <usama.arif@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1149
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>

4 files changed, 679 insertions, 34 deletions

diff --git a/tests/validation/NEON/Cast.cpp b/tests/validation/NEON/Cast.cpp
new file mode 100644
index 0000000000..9300ad4163
--- /dev/null
+++ b/tests/validation/NEON/Cast.cpp
@@ -0,0 +1,186 @@
+/*
+ * Copyright (c) 2019 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/NEON/functions/NECast.h"
+#include "arm_compute/runtime/Tensor.h"
+#include "arm_compute/runtime/TensorAllocator.h"
+#include "tests/NEON/Accessor.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/CastFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+// Tolerance
+constexpr AbsoluteTolerance<float> one_tolerance(1);
+constexpr AbsoluteTolerance<float> zero_tolerance(0);
+
+/*
+ *This function ignores the scale and zeroPoint of quanized tensors,so QASYMM8 input is treated as uint8 values.
+*/
+
+/** Input data sets **/
+// QASYMM8
+const auto CastQASYMM8toF16Dataset     = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::F16));
+const auto CastQASYMM8toF32Dataset     = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::F32));
+const auto CastQASYMM8toS32Dataset     = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::S32));
+
+// U8
+const auto CastU8toU16Dataset = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U16));
+const auto CastU8toS16Dataset = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S16));
+const auto CastU8toS32Dataset = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S32));
+const auto CastU8toF32Dataset = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::F32));
+
+// U16
+const auto CastU16toU8Dataset  = combine(framework::dataset::make("DataType", DataType::U16), framework::dataset::make("DataType", DataType::U8));
+const auto CastU16toU32Dataset = combine(framework::dataset::make("DataType", DataType::U16), framework::dataset::make("DataType", DataType::U32));
+
+// S16
+const auto CastS16toU8Dataset  = combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::U8));
+const auto CastS16toS32Dataset = combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::S32));
+
+//S32
+const auto CastS32toF16Dataset     = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::F16));
+const auto CastS32toU8Dataset     = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::U8));
+const auto CastS32toF32Dataset     = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::F32));
+const auto CastS32toQASYMM8Dataset     = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::QASYMM8));
+
+// F16
+const auto CastF16toF32Dataset     = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F32));
+const auto CastF16toS32Dataset     = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::S32));
+const auto CastF16toQASYMM8Dataset     = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::QASYMM8));
+
+
+// F32
+const auto CastF32toU8Dataset     = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::U8));
+const auto CastF32toF16Dataset     = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F16));
+const auto CastF32toS32Dataset     = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::S32));
+const auto CastF32toQASYMM8Dataset     = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::QASYMM8));
+
+} // namespace
+
+TEST_SUITE(NEON)
+TEST_SUITE(Cast)
+template <typename T>
+using NECastToU8Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint8_t>;
+template <typename T>
+using NECastToU16Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint16_t>;
+template <typename T>
+using NECastToS16Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, int16_t>;
+template <typename T>
+using NECastToU32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint32_t>;
+template <typename T>
+using NECastToS32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, int32_t>;
+template <typename T>
+using NECastToF16Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, half>;
+template <typename T>
+using NECastToF32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, float>;
+template <typename T>
+using NECastToQASYMM8Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint8_t>;
+
+#define CAST_SUITE(NAME, idt, odt, type, dataset, tolerance)                                                                     \
+    TEST_SUITE(NAME)                                                                                                             \
+    DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), datasets::ConvertPolicies()),    \
+                   shape, policy)                                                                                                \
+    {                                                                                                                            \
+        Tensor src = create_tensor<Tensor>(shape, idt, 1);                                                                       \
+        Tensor dst = create_tensor<Tensor>(shape, odt, 1);                                                                       \
+        \
+        NECast cast;                                                                                                             \
+        cast.configure(&src, &dst, policy);                                                                                      \
+        \
+        const ValidRegion valid_region = shape_to_valid_region(shape);                                                           \
+        validate(dst.info()->valid_region(), valid_region);                                                                      \
+        \
+        const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();                                         \
+        validate(src.info()->padding(), padding);                                                                                \
+        validate(dst.info()->padding(), padding);                                                                                \
+    }                                                                                                                            \
+    FIXTURE_DATA_TEST_CASE(RunSmall, type, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), dataset), \
+                                                                                      datasets::ConvertPolicies()))              \
+    {                                                                                                                            \
+        validate(Accessor(_target), _reference, tolerance);                                                                      \
+    }                                                                                                                            \
+    TEST_SUITE_END()
+
+//QASYMM8
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(QASYMM8_to_F16, DataType::QASYMM8, DataType::F16, NECastToF16Fixture<uint8_t>, CastQASYMM8toF16Dataset, one_tolerance)
+#endif //  __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(QASYMM8_to_F32, DataType::QASYMM8, DataType::F32, NECastToF32Fixture<uint8_t>, CastQASYMM8toF32Dataset, one_tolerance)
+CAST_SUITE(QASYMM8_to_S32, DataType::QASYMM8, DataType::S32, NECastToS32Fixture<uint8_t>, CastQASYMM8toS32Dataset, one_tolerance)
+
+// U8
+CAST_SUITE(U8_to_U16, DataType::U8, DataType::U16, NECastToU16Fixture<uint8_t>, CastU8toU16Dataset, zero_tolerance)
+CAST_SUITE(U8_to_S16, DataType::U8, DataType::S16, NECastToS16Fixture<uint8_t>, CastU8toS16Dataset, zero_tolerance)
+CAST_SUITE(U8_to_S32, DataType::U8, DataType::S32, NECastToS32Fixture<uint8_t>, CastU8toS32Dataset, zero_tolerance)
+CAST_SUITE(U8_to_F32, DataType::U8, DataType::F32, NECastToF32Fixture<uint8_t>, CastU8toF32Dataset, zero_tolerance)
+
+// U16
+CAST_SUITE(U16_to_U8, DataType::U16, DataType::U8, NECastToU8Fixture<uint16_t>, CastU16toU8Dataset, zero_tolerance)
+CAST_SUITE(U16_to_U32, DataType::U16, DataType::U32, NECastToU32Fixture<uint16_t>, CastU16toU32Dataset, zero_tolerance)
+
+// S16
+CAST_SUITE(S16_to_U8, DataType::S16, DataType::U8, NECastToU8Fixture<int16_t>, CastS16toU8Dataset, zero_tolerance)
+CAST_SUITE(S16_to_S32, DataType::S16, DataType::S32, NECastToS32Fixture<int16_t>, CastS16toS32Dataset, zero_tolerance)
+
+// S32
+CAST_SUITE(S32_to_QASYMM8, DataType::S32, DataType::QASYMM8, NECastToQASYMM8Fixture<int32_t>, CastS32toQASYMM8Dataset, one_tolerance)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(S32_to_F16, DataType::S32, DataType::F16, NECastToF16Fixture<int32_t>, CastS32toF16Dataset, zero_tolerance)
+#endif //  __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(S32_to_F32, DataType::S32, DataType::F32, NECastToF32Fixture<int32_t>, CastS32toF32Dataset, one_tolerance)
+CAST_SUITE(S32_to_U8, DataType::S32, DataType::U8, NECastToU8Fixture<int32_t>, CastS32toU8Dataset, one_tolerance)
+
+// F16
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(F16_to_QASYMM8, DataType::F16, DataType::QASYMM8, NECastToQASYMM8Fixture<half>, CastF16toQASYMM8Dataset, one_tolerance)
+CAST_SUITE(F16_to_F32, DataType::F16, DataType::F32, NECastToF32Fixture<half>, CastF16toF32Dataset, zero_tolerance)
+CAST_SUITE(F16_to_S32, DataType::F16, DataType::S32, NECastToS32Fixture<half>, CastF16toS32Dataset, one_tolerance)
+#endif //  __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+// F32
+CAST_SUITE(F32_to_QASYMM8, DataType::F32, DataType::QASYMM8, NECastToQASYMM8Fixture<float>, CastF32toQASYMM8Dataset, one_tolerance)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(F32_to_F16, DataType::F32, DataType::F16, NECastToF16Fixture<float>, CastF32toF16Dataset, zero_tolerance)
+#endif //  __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+CAST_SUITE(F32_to_S32, DataType::F32, DataType::S32, NECastToS32Fixture<float>, CastF32toS32Dataset, one_tolerance)
+CAST_SUITE(F32_to_U8, DataType::F32, DataType::S32, NECastToS32Fixture<float>, CastF32toS32Dataset, one_tolerance)
+
+TEST_SUITE_END() // Cast
+TEST_SUITE_END() // NEON
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/NEON/DepthConvertLayer.cpp b/tests/validation/NEON/DepthConvertLayer.cpp
index cee7bdedc2..2592c5d5a1 100644
--- a/tests/validation/NEON/DepthConvertLayer.cpp
+++ b/tests/validation/NEON/DepthConvertLayer.cpp
@@ -46,15 +46,28 @@ namespace
 /** Input data sets **/
 const auto DepthConvertLayerQASYMM8toF16Dataset   = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::F16));
 const auto DepthConvertLayerQASYMM8toF32Dataset   = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::F32));
+const auto DepthConvertLayerQASYMM8toS32Dataset   = combine(framework::dataset::make("DataType", DataType::QASYMM8), framework::dataset::make("DataType", DataType::S32));
 const auto DepthConvertLayerU8toU16Dataset        = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::U16));
 const auto DepthConvertLayerU8toS16Dataset        = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S16));
 const auto DepthConvertLayerU8toS32Dataset        = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::S32));
+const auto DepthConvertLayerU8toF16Dataset        = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::F16));
+const auto DepthConvertLayerU8toF32Dataset        = combine(framework::dataset::make("DataType", DataType::U8), framework::dataset::make("DataType", DataType::F32));
 const auto DepthConvertLayerU16toU8Dataset        = combine(framework::dataset::make("DataType", DataType::U16), framework::dataset::make("DataType", DataType::U8));
 const auto DepthConvertLayerU16toU32Dataset       = combine(framework::dataset::make("DataType", DataType::U16), framework::dataset::make("DataType", DataType::U32));
 const auto DepthConvertLayerS16toU8Dataset        = combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::U8));
 const auto DepthConvertLayerS16toS32Dataset       = combine(framework::dataset::make("DataType", DataType::S16), framework::dataset::make("DataType", DataType::S32));
+const auto DepthConvertLayerF16toU8Dataset       = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::U8));
 const auto DepthConvertLayerF16toF32Dataset       = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F32));
+const auto DepthConvertLayerF16toS32Dataset       = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::S32));
 const auto DepthConvertLayerF32toF16Dataset       = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F16));
+const auto DepthConvertLayerF32toS32Dataset       = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::S32));
+const auto DepthConvertLayerF32toU8Dataset       = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::U8));
+
+const auto DepthConvertLayerS32toF32Dataset       = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::F32));
+const auto DepthConvertLayerS32toQASYMM8Dataset       = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::QASYMM8));
+const auto DepthConvertLayerS32toF16Dataset       = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::F16));
+const auto DepthConvertLayerS32toU8Dataset       = combine(framework::dataset::make("DataType", DataType::S32), framework::dataset::make("DataType", DataType::U8));
+
 const auto DepthConvertLayerF16toQASYMM8Dataset   = combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::QASYMM8));
 const auto DepthConvertLayerF32toQASYMM8Dataset   = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::QASYMM8));
 const auto DepthConvertLayerShiftDatasetNightly   = framework::dataset::make("Shift", 0, 7);
@@ -62,6 +75,8 @@ const auto DepthConvertLayerShiftDatasetPrecommit = framework::dataset::make("Sh
 const auto DepthConvertLayerZeroShiftDataset      = framework::dataset::make("Shift", 0);
 
 constexpr AbsoluteTolerance<uint8_t> tolerance_qasymm8(1);
+constexpr AbsoluteTolerance<int32_t> tolerance_one_int32(1);
+constexpr AbsoluteTolerance<uint8_t> tolerance_one_uint8(1);
 } // namespace
 
 TEST_SUITE(NEON)
@@ -70,23 +85,15 @@ TEST_SUITE(DepthConvertLayer)
 // *INDENT-OFF*
 // clang-format off
 DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
-               framework::dataset::make("InputInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::QASYMM8),  // Invalid data type combination
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),  // Invalid data type combination
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U16),  // Invalid data type combination
+               framework::dataset::make("InputInfo", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U16),  // Invalid data type combination
                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),  // Invalid data type combination
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),  // Invalid data type combination
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F16),  // Invalid data type combination
                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),  // Mismatching shapes
                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),  // Invalid shift
                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),  // Valid
                                                      }),
-               framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S16),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+               framework::dataset::make("OutputInfo",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
                                                        TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
-                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
                                                      })),
                framework::dataset::make("Policy",{ ConvertPolicy::WRAP,
@@ -99,7 +106,7 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(
                                                      })),
                framework::dataset::make("Shift",{ 1, 1, 1, 1, 1, 1, 8, 1,
                                                      })),
-               framework::dataset::make("Expected", { false, false, false, false, false, false, false, true})),
+               framework::dataset::make("Expected", { false, false, false, false, true})),
                input_info, output_info, policy, shift, expected)
 {
     ARM_COMPUTE_EXPECT(bool(NEDepthConvertLayer::validate(&input_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), policy, shift)) == expected, framework::LogLevel::ERRORS);
@@ -127,6 +134,8 @@ template <typename T>
 using NEDepthConvertLayerQuantizedToF16Fixture = DepthConvertLayerValidationQuantizedFixture<Tensor, Accessor, NEDepthConvertLayer, T, half>;
 template <typename T>
 using NEDepthConvertLayerQuantizedToF32Fixture = DepthConvertLayerValidationQuantizedFixture<Tensor, Accessor, NEDepthConvertLayer, T, float>;
+template <typename T>
+using NEDepthConvertLayerQuantizedToS32Fixture = DepthConvertLayerValidationQuantizedFixture<Tensor, Accessor, NEDepthConvertLayer, T, int32_t>;
 
 TEST_SUITE(QASYMM8_to_F32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
@@ -171,6 +180,49 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerQuantizedToF32Fixture<uint8_
 }
 TEST_SUITE_END() // QASYMM8_to_F32
 
+TEST_SUITE(QASYMM8_to_S32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::QASYMM8, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::S32, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerQuantizedToS32Fixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                       DepthConvertLayerQASYMM8toS32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                       DepthConvertLayerZeroShiftDataset),
+                       framework::dataset::make("QuantizationInfo", { QuantizationInfo(0.5f, 10) })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerQuantizedToS32Fixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(combine(datasets::LargeShapes(),
+                       DepthConvertLayerQASYMM8toS32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                       DepthConvertLayerZeroShiftDataset),
+                       framework::dataset::make("QuantizationInfo", { QuantizationInfo(0.5f, 10) })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // QASYMM8_to_S32
+
 TEST_SUITE(U8_to_U16)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                    DepthConvertLayerShiftDatasetNightly),
@@ -290,6 +342,90 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToS32Fixture<uint8_t>, frame
 }
 TEST_SUITE_END() // U8_to_S32
 
+
+TEST_SUITE(U8_to_F32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerShiftDatasetNightly),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::U8, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::F32, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToF32Fixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerU8toF32Dataset),
+                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                      DepthConvertLayerShiftDatasetPrecommit))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF32Fixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerU8toF32Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerShiftDatasetNightly))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // U8_to_F32
+
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+TEST_SUITE(U8_to_F16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerShiftDatasetNightly),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::U8, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::F16, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToF16Fixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerU8toF16Dataset),
+                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                      DepthConvertLayerShiftDatasetPrecommit))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF16Fixture<uint8_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerU8toF16Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerShiftDatasetNightly))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // U8_to_F36
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+
 TEST_SUITE(U16_to_U8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                    DepthConvertLayerShiftDatasetNightly),
@@ -490,6 +626,48 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToQASYMM8Fixture<half>, fram
 }
 TEST_SUITE_END() // F16_to_QASYMM8
 
+
+TEST_SUITE(F16_to_U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::F16, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::U8, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToU8Fixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerF16toU8Dataset),
+                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                   DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_uint8);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToU8Fixture<half>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerF16toU8Dataset),
+                                                                                                                 framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                 DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_uint8);
+}
+TEST_SUITE_END() // F16_to_U8
+
+
+
 TEST_SUITE(F16_to_F32)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                    DepthConvertLayerZeroShiftDataset),
@@ -529,6 +707,46 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF32Fixture<half>, framewor
 }
 TEST_SUITE_END() // F16_to_F32
 
+TEST_SUITE(F16_to_S32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::F16, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::S32, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToS32Fixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerF16toS32Dataset),
+                                                                                                                   framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                   DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToS32Fixture<half>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerF16toS32Dataset),
+                                                                                                                 framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                 DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+
+TEST_SUITE_END() // F16_to_S32
+
 TEST_SUITE(QASYMM8_to_F16)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                    DepthConvertLayerZeroShiftDataset),
@@ -610,8 +828,128 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF16Fixture<float>, framewo
     validate(Accessor(_target), _reference);
 }
 TEST_SUITE_END() // F32_to_F16
+
+TEST_SUITE(S32_to_F16)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::S32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::F16, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToF16Fixture<int32_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerS32toF16Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF16Fixture<int32_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerS32toF16Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // S32_to_F16
+
 #endif           /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
 
+TEST_SUITE(F32_to_S32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::F32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::S32, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToS32Fixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerF32toS32Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToS32Fixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerF32toS32Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+TEST_SUITE_END() // F32_to_S32
+
+TEST_SUITE(F32_to_U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::F32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::U8, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToU8Fixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerF32toU8Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToU8Fixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerF32toU8Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_one_int32);
+}
+TEST_SUITE_END() // F32_to_U8
+
+
+
 TEST_SUITE(F32_to_QASYMM8)
 DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                    DepthConvertLayerZeroShiftDataset),
@@ -654,6 +992,132 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToQASYMM8Fixture<float>, fra
     validate(Accessor(_target), _reference, tolerance_qasymm8);
 }
 TEST_SUITE_END() // F32_to_QASYMM8
+
+
+TEST_SUITE(S32_to_F32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::S32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::F32, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToF32Fixture<int32_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerS32toF32Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToF32Fixture<int32_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerS32toF32Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // S32_to_F32
+
+TEST_SUITE(S32_to_QASYMM8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::S32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::QASYMM8, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToQASYMM8Fixture<int32_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+                                                                                                                        DepthConvertLayerS32toQASYMM8Dataset),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                                                                                                                        DepthConvertLayerZeroShiftDataset),
+                                                                                                                        framework::dataset::make("QuantizationInfo", { QuantizationInfo(0.5f, 10) })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_qasymm8);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToQASYMM8Fixture<int32_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(combine(datasets::LargeShapes(),
+                                                                                                                      DepthConvertLayerS32toQASYMM8Dataset),
+                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                                                                                                                      DepthConvertLayerZeroShiftDataset),
+                                                                                                                      framework::dataset::make("QuantizationInfo", { QuantizationInfo(0.5f, 10) })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_qasymm8);
+}
+TEST_SUITE_END() // S32_to_QASYMM8
+
+TEST_SUITE(S32_to_U8)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                   DepthConvertLayerZeroShiftDataset),
+               shape, policy, shift)
+{
+    // Create tensors
+    Tensor src = create_tensor<Tensor>(shape, DataType::S32, 1);
+    Tensor dst = create_tensor<Tensor>(shape, DataType::U8, 1);
+
+    // Create and Configure function
+    NEDepthConvertLayer depth_convert;
+    depth_convert.configure(&src, &dst, policy, shift);
+
+    // Validate valid region
+    const ValidRegion valid_region = shape_to_valid_region(shape);
+    validate(dst.info()->valid_region(), valid_region);
+
+    // Validate padding
+    const PaddingSize padding = PaddingCalculator(shape.x(), 16).required_padding();
+    validate(src.info()->padding(), padding);
+    validate(dst.info()->padding(), padding);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEDepthConvertLayerToU8Fixture<int32_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), DepthConvertLayerS32toU8Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                    DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+FIXTURE_DATA_TEST_CASE(RunLarge, NEDepthConvertLayerToU8Fixture<int32_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), DepthConvertLayerS32toU8Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  DepthConvertLayerZeroShiftDataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // S32_to_U8
+
+
+
+
 TEST_SUITE_END() // DepthConvertLayer
 TEST_SUITE_END() // NEON
 } // namespace validation
diff --git a/tests/validation/fixtures/DepthConvertLayerFixture.h b/tests/validation/fixtures/DepthConvertLayerFixture.h
index a3d379eede..3fe12709c8 100644
--- a/tests/validation/fixtures/DepthConvertLayerFixture.h
+++ b/tests/validation/fixtures/DepthConvertLayerFixture.h
@@ -40,6 +40,7 @@ namespace test
 {
 namespace validation
 {
+/*  This function ignores the scale and zeroPoint of quanized tensors, i.e. QASYMM8 input is treated as uint8 values.*/
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T1, typename T2>
 class DepthConvertLayerValidationBaseFixture : public framework::Fixture
 {
@@ -55,7 +56,7 @@ public:
 
 protected:
     template <typename U>
-    void fill(U &&tensor, int i)
+    void fill(U &&tensor, int i, DataType dt_in, DataType dt_out)
     {
         if(is_data_type_quantized(tensor.data_type()))
         {
@@ -66,7 +67,16 @@ protected:
         }
         else
         {
-            library->fill_tensor_uniform(tensor, i);
+            // When converting S32 to F16, both reference and NEON implementations are + or - infinity outside the F16 range.
+            if(dt_in==DataType::S32 && dt_out==DataType::F16)
+            {
+                std::uniform_int_distribution<int32_t> distribution_s32(-65504, 65504);
+                library->fill(tensor, distribution_s32, i);
+            }
+            else
+            {
+                library->fill_tensor_uniform(tensor, i);
+            }
         }
     }
 
@@ -91,7 +101,7 @@ protected:
         ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
 
         // Fill tensors
-        fill(AccessorType(src), 0);
+        fill(AccessorType(src), 0, dt_in, dt_out);
 
         // Compute function
         depth_convert.run();
@@ -105,7 +115,7 @@ protected:
         SimpleTensor<T1> src{ shape, dt_in, 1, _quantization_info };
 
         // Fill reference
-        fill(src, 0);
+        fill(src, 0, dt_in, dt_out);
 
         return reference::depth_convert<T1, T2>(src, dt_out, policy, shift);
     }
diff --git a/tests/validation/reference/DepthConvertLayer.cpp b/tests/validation/reference/DepthConvertLayer.cpp
index 6d9f98db5a..7da0011fbb 100644
--- a/tests/validation/reference/DepthConvertLayer.cpp
+++ b/tests/validation/reference/DepthConvertLayer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2019 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -48,14 +48,7 @@ SimpleTensor<T2> depth_convert(const SimpleTensor<T1> &src, DataType dt_out, Con
     {
         for(int i = 0; i < src.num_elements(); ++i)
         {
-            if(is_data_type_quantized(src.data_type()))
-            {
-                result[i] = scvt_f32_qasymm8(src[i], src.quantization_info().scale, src.quantization_info().offset);
-            }
-            else
-            {
-                result[i] = src[i] << shift;
-            }
+            result[i] = src[i] << shift;
         }
     }
     // Down-casting
@@ -82,16 +75,8 @@ SimpleTensor<T2> depth_convert(const SimpleTensor<T1> &src, DataType dt_out, Con
         // Always saturate on floats
         for(int i = 0; i < src.num_elements(); ++i)
         {
-            if(is_data_type_quantized(dt_out))
-            {
-                T1 val    = utils::rounding::round_half_away_from_zero(src[i]);
-                result[i] = sqcvt_qasymm8_f32(val, src.quantization_info().scale, src.quantization_info().offset);
-            }
-            else
-            {
-                T1 val    = utils::rounding::round_half_away_from_zero(src[i]);
-                result[i] = utils::cast::saturate_cast<T2>(val);
-            }
+            T1 val    = utils::rounding::round_half_away_from_zero(src[i]);
+            result[i] = utils::cast::saturate_cast<T2>(val);
         }
     }
     else