Add support for input S64/U64 in CpuCastKernel

* The kernel now supports the following conversions:
    S64 -> F32
    U64 -> F32

* Resolves MLCE-1089

Change-Id: I277cf58b78d919fde25947520d2056e1412c7f82
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9935
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Viet-Hoa Do <viet-hoa.do@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

3 files changed, 23 insertions, 0 deletions

diff --git a/tests/SimpleTensor.h b/tests/SimpleTensor.h
index 9ea171d492..419621e808 100644
--- a/tests/SimpleTensor.h
+++ b/tests/SimpleTensor.h
@@ -392,6 +392,8 @@ int SimpleTensor<T>::num_channels() const
         case Format::S16:
         case Format::U32:
         case Format::S32:
+        case Format::U64:
+        case Format::S64:
         case Format::F16:
         case Format::F32:
             return 1;
diff --git a/tests/validation/NEON/Cast.cpp b/tests/validation/NEON/Cast.cpp
index 166847ed66..a1ddcc9cad 100644
--- a/tests/validation/NEON/Cast.cpp
+++ b/tests/validation/NEON/Cast.cpp
@@ -101,6 +101,11 @@ const auto CastF32toS32Dataset            = combine(framework::dataset::make("Da
 const auto CastF32toQASYMM8Dataset        = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::QASYMM8));
 const auto CastF32toQASYMM8_SIGNEDDataset = combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::QASYMM8_SIGNED));
 
+// U64
+const auto CastU64toF32Dataset = combine(framework::dataset::make("DataType", DataType::U64), framework::dataset::make("DataType", DataType::F32));
+
+// S64
+const auto CastS64toF32Dataset = combine(framework::dataset::make("DataType", DataType::S64), framework::dataset::make("DataType", DataType::F32));
 } // namespace
 
 TEST_SUITE(NEON)
@@ -108,6 +113,8 @@ TEST_SUITE(Cast)
 template <typename T>
 using NECastToU8Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint8_t>;
 template <typename T>
+using NECastToS8Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, int8_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>;
@@ -116,6 +123,10 @@ using NECastToU32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, ui
 template <typename T>
 using NECastToS32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, int32_t>;
 template <typename T>
+using NECastToU64Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, uint64_t>;
+template <typename T>
+using NECastToS64Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, int64_t>;
+template <typename T>
 using NECastToF16Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, half>;
 template <typename T>
 using NECastToF32Fixture = CastValidationFixture<Tensor, Accessor, NECast, T, float>;
@@ -189,6 +200,14 @@ CAST_SUITE(F32_to_F16, DataType::F32, DataType::F16, NECastToF16Fixture<float>,
 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)
 
+#ifdef __aarch64__
+// S64
+CAST_SUITE(S64_to_F32, DataType::S64, DataType::F32, NECastToF32Fixture<int64_t>, CastS64toF32Dataset, zero_tolerance)
+
+// U64
+CAST_SUITE(U64_to_F32, DataType::U64, DataType::F32, NECastToF32Fixture<uint64_t>, CastU64toF32Dataset, zero_tolerance)
+#endif // __aarch64__
+
 DATA_TEST_CASE(KernelSelectionDstFP16, framework::DatasetMode::ALL,
                combine(framework::dataset::make("CpuExt", std::string("NEON")),
                        framework::dataset::make("DataType",
diff --git a/tests/validation/reference/DepthConvertLayer.cpp b/tests/validation/reference/DepthConvertLayer.cpp
index 8797722f00..1e4939129e 100644
--- a/tests/validation/reference/DepthConvertLayer.cpp
+++ b/tests/validation/reference/DepthConvertLayer.cpp
@@ -194,6 +194,7 @@ template SimpleTensor<int16_t> depth_convert(const SimpleTensor<int64_t> &src, D
 template SimpleTensor<uint32_t> depth_convert(const SimpleTensor<int64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 template SimpleTensor<int32_t> depth_convert(const SimpleTensor<int64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 template SimpleTensor<half> depth_convert(const SimpleTensor<int64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
+template SimpleTensor<float> depth_convert(const SimpleTensor<int64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 
 // U64
 template SimpleTensor<uint8_t> depth_convert(const SimpleTensor<uint64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
@@ -203,6 +204,7 @@ template SimpleTensor<int16_t> depth_convert(const SimpleTensor<uint64_t> &src,
 template SimpleTensor<uint32_t> depth_convert(const SimpleTensor<uint64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 template SimpleTensor<int32_t> depth_convert(const SimpleTensor<uint64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 template SimpleTensor<half> depth_convert(const SimpleTensor<uint64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
+template SimpleTensor<float> depth_convert(const SimpleTensor<uint64_t> &src, DataType dt_out, ConvertPolicy policy, uint32_t shift);
 } // namespace reference
 } // namespace validation
 } // namespace test