COMPMID-1768: NEON: Implement RealDiv
Change-Id: I0868669f7b733df141794fba1d79436e7581bd3a
Reviewed-on: https://review.mlplatform.org/426
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h b/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
index f02f71b..1271da7 100644
--- a/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
+++ b/arm_compute/core/NEON/kernels/NEElementwiseOperationKernel.h
@@ -94,10 +94,8 @@
class NEArithmeticOperationKernel : public NEElementwiseOperationKernel
{
public:
- NEArithmeticOperationKernel()
- : NEElementwiseOperationKernel()
- {
- }
+ /** Default constructor */
+ NEArithmeticOperationKernel() = default;
/** Static function to check if given info will lead to a valid configuration of @ref NEArithmeticOperationKernel
*
@@ -126,13 +124,40 @@
static Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, const ITensorInfo &output);
};
+class NEDivisionOperationKernel : public NEArithmeticOperationKernel
+{
+public:
+ /** Default constructor */
+ NEDivisionOperationKernel() = default;
+
+ /** Static function to check if given info will lead to a valid configuration of @ref NEArithmeticOperationKernel
+ *
+ * @param[in] input1 First tensor input. Data types supported: F16/F32.
+ * @param[in] input2 Second tensor input. Data types supported: Same as @p input1.
+ * @param[in] output Output tensor. Data types supported: Same as @p input1.
+ */
+ void configure(const ITensor *input1, const ITensor *input2, ITensor *output);
+
+ /** Static function to check if given info will lead to a valid configuration of @ref NEArithmeticOperationKernel
+ *
+ * @param[in] input1 First tensor input info. Data types supported: F16/F32.
+ * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
+ * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ *
+ * @return a Status
+ */
+ static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+
+protected:
+ // Inherited methods overridden:
+ static Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, const ITensorInfo &output);
+};
+
class NEComparisonOperationKernel : public NEElementwiseOperationKernel
{
public:
- NEComparisonOperationKernel()
- : NEElementwiseOperationKernel()
- {
- }
+ /** Default constructor */
+ NEComparisonOperationKernel() = default;
/** Static function to check if given info will lead to a valid configuration of @ref NEComparisonOperationKernel
*
diff --git a/arm_compute/core/NEON/wrapper/intrinsics/div.h b/arm_compute/core/NEON/wrapper/intrinsics/div.h
new file mode 100644
index 0000000..d9f80d0
--- /dev/null
+++ b/arm_compute/core/NEON/wrapper/intrinsics/div.h
@@ -0,0 +1,73 @@
+/*
+ * 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.
+ */
+#ifndef __ARM_COMPUTE_WRAPPER_DIV_H__
+#define __ARM_COMPUTE_WRAPPER_DIV_H__
+
+#include "arm_compute/core/NEON/NEMath.h"
+#include <arm_neon.h>
+
+namespace arm_compute
+{
+namespace wrapper
+{
+#ifdef __aarch64__
+
+#define VDIV_IMPL(stype, vtype, prefix, postfix) \
+ inline vtype vdiv(const vtype &a, const vtype &b) \
+ { \
+ return prefix##_##postfix(a, b); \
+ }
+VDIV_IMPL(float32x2_t, float32x2_t, vdiv, f32)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+VDIV_IMPL(float16x4_t, float16x4_t, vdiv, f16)
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+VDIV_IMPL(float32x4_t, float32x4_t, vdivq, f32)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+VDIV_IMPL(float16x8_t, float16x8_t, vdivq, f16)
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+#else // __aarch64__
+
+#define VDIV_IMPL(stype, vtype, mul_prefix, inv_prefix, postfix) \
+ inline vtype vdiv(const vtype &a, const vtype &b) \
+ { \
+ return mul_prefix##_##postfix(a, inv_prefix##_##postfix(b)); \
+ }
+VDIV_IMPL(float32x2_t, float32x2_t, vmul, vinv, f32)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+VDIV_IMPL(float16x4_t, float16x4_t, vmul, vinv, f16)
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+VDIV_IMPL(float32x4_t, float32x4_t, vmulq, vinvq, f32)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+VDIV_IMPL(float16x8_t, float16x8_t, vmulq, vinvq, f16)
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
+#endif // __aarch64__
+
+#undef VDIV_IMPL
+} // namespace wrapper
+} // namespace arm_compute
+#endif /* __ARM_COMPUTE_WRAPPER_DIV_H__ */
diff --git a/arm_compute/core/NEON/wrapper/intrinsics/intrinsics.h b/arm_compute/core/NEON/wrapper/intrinsics/intrinsics.h
index c8f4a6e..012f686 100644
--- a/arm_compute/core/NEON/wrapper/intrinsics/intrinsics.h
+++ b/arm_compute/core/NEON/wrapper/intrinsics/intrinsics.h
@@ -33,6 +33,7 @@
#include "arm_compute/core/NEON/wrapper/intrinsics/cgt.h"
#include "arm_compute/core/NEON/wrapper/intrinsics/clt.h"
#include "arm_compute/core/NEON/wrapper/intrinsics/combine.h"
+#include "arm_compute/core/NEON/wrapper/intrinsics/div.h"
#include "arm_compute/core/NEON/wrapper/intrinsics/dup_n.h"
#include "arm_compute/core/NEON/wrapper/intrinsics/exp.h"
#include "arm_compute/core/NEON/wrapper/intrinsics/gethigh.h"
diff --git a/arm_compute/runtime/NEON/functions/NEElementwiseOperations.h b/arm_compute/runtime/NEON/functions/NEElementwiseOperations.h
index cd9ed24..ca3717a 100644
--- a/arm_compute/runtime/NEON/functions/NEElementwiseOperations.h
+++ b/arm_compute/runtime/NEON/functions/NEElementwiseOperations.h
@@ -109,7 +109,33 @@
static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
};
-/** Basic function to run @ref NEComparisonOperationKernel
+/** Basic function to run @ref NEArithmeticOperationKernel for division
+ *
+ * @note The tensor data type for the inputs must be F16/F32.
+ * @note The function performs a squared different operation between two tensors (i.e., out[i] = in1[i] / in2[i])
+ */
+class NEElementwiseDivision : public INESimpleFunction
+{
+public:
+ /** Initialise the kernel's inputs, output and conversion policy.
+ *
+ * @param[in, out] input1 First tensor input. Data types supported: F16/F32.
+ * @param[in, out] input2 Second tensor input. Data types supported: Same as @p input1.
+ * @param[out] output Output tensor. Data types supported: Same as @p input1.
+ */
+ void configure(ITensor *input1, ITensor *input2, ITensor *output);
+ /** Static function to check if given info will lead to a valid configuration of @ref NEArithmeticOperationKernel for division
+ *
+ * @param[in] input1 First tensor input info. Data types supported: F16/F32.
+ * @param[in] input2 Second tensor input info. Data types supported: Same as @p input1.
+ * @param[in] output Output tensor info. Data types supported: Same as @p input1.
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output);
+};
+
+/** Basic function to run @ref NEComparisonOperationKernel.
*
* @note The tensor data type for the inputs must be QASYMM8/S16/F16/S32/F32.
* @note The function performs a comparison operation between two tensors.
diff --git a/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp b/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp
index 88fd730..99a3b5a 100644
--- a/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp
+++ b/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp
@@ -123,6 +123,11 @@
res = (a - b) * (a - b);
break;
}
+ case ArithmeticOperation::DIV:
+ {
+ res = a / b;
+ break;
+ }
default:
ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
}
@@ -154,7 +159,6 @@
res = wrapper::vmul(tmp, tmp);
break;
}
-
default:
ARM_COMPUTE_ERROR("NOT_SUPPORTED!");
}
@@ -162,6 +166,20 @@
return res;
}
+template <>
+inline float32x4_t elementwise_arithm_op<ArithmeticOperation::DIV, float32x4_t>(const float32x4_t &a, const float32x4_t &b)
+{
+ return wrapper::vdiv(a, b);
+}
+
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+template <>
+inline float16x8_t elementwise_arithm_op<ArithmeticOperation::DIV, float16x8_t>(const float16x8_t &a, const float16x8_t &b)
+{
+ return wrapper::vdiv(a, b);
+}
+#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+
template <ArithmeticOperation op>
inline float32x4x4_t elementwise_arithm_op(const float32x4x4_t &a, const float32x4x4_t &b)
{
@@ -833,6 +851,28 @@
return Status{};
}
+/** The division operator */
+
+void NEDivisionOperationKernel::configure(const ITensor *input1, const ITensor *input2, ITensor *output)
+{
+ ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*input1->info(), *input2->info(), *output->info()));
+ configure_common(input1, input2, output);
+ _function = configure_arithm_func<ArithmeticOperation::DIV>(input1, input2, output);
+}
+
+Status NEDivisionOperationKernel::validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, const ITensorInfo &output)
+{
+ ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input1, 1, DataType::F16, DataType::F32);
+ return NEArithmeticOperationKernel::validate_arguments(input1, input2, output);
+}
+
+Status NEDivisionOperationKernel::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+{
+ ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1, input2, output);
+ ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(*input1, *input2, *output));
+ return Status{};
+}
+
/** Comparison operators (equal, not equal, less than, greater than, less than or equal, greater than or equal) */
void NEComparisonOperationKernel::configure(ComparisonOperation op, const ITensor *input1, const ITensor *input2, ITensor *output)
diff --git a/src/runtime/NEON/functions/NEElementwiseOperators.cpp b/src/runtime/NEON/functions/NEElementwiseOperators.cpp
index 711e99e..4c57068 100644
--- a/src/runtime/NEON/functions/NEElementwiseOperators.cpp
+++ b/src/runtime/NEON/functions/NEElementwiseOperators.cpp
@@ -67,6 +67,18 @@
return NEArithmeticOperationKernel::validate(ArithmeticOperation::SQUARED_DIFF, input1, input2, output);
}
+void NEElementwiseDivision::configure(ITensor *input1, ITensor *input2, ITensor *output)
+{
+ auto k = arm_compute::support::cpp14::make_unique<NEDivisionOperationKernel>();
+ k->configure(input1, input2, output);
+ _kernel = std::move(k);
+}
+
+Status NEElementwiseDivision::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output)
+{
+ return NEDivisionOperationKernel::validate(input1, input2, output);
+}
+
template <ComparisonOperation COP>
void NEElementwiseComparisonStatic<COP>::configure(ITensor *input1, ITensor *input2, ITensor *output)
{
diff --git a/tests/validation/NEON/ElementwiseDivision.cpp b/tests/validation/NEON/ElementwiseDivision.cpp
new file mode 100644
index 0000000..86a2056
--- /dev/null
+++ b/tests/validation/NEON/ElementwiseDivision.cpp
@@ -0,0 +1,155 @@
+/*
+ * 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/NEElementwiseOperations.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/ShapeDatasets.h"
+#include "tests/framework/Asserts.h"
+#include "tests/framework/Macros.h"
+#include "tests/framework/datasets/Datasets.h"
+#include "tests/validation/Validation.h"
+#include "tests/validation/fixtures/ElementwiseOperationsFixture.h"
+
+namespace arm_compute
+{
+namespace test
+{
+namespace validation
+{
+namespace
+{
+RelativeTolerance<float> tolerance_fp32(0.000001f);
+/** Input data sets **/
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+RelativeTolerance<half> tolerance_fp16(static_cast<half>(0.01f));
+const auto ElementwiseDivisionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
+ framework::dataset::make("DataType", DataType::F16));
+#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+const auto ElementwiseDivisionFP32Dataset = combine(combine(framework::dataset::make("DataType", DataType::F32), framework::dataset::make("DataType", DataType::F32)),
+ framework::dataset::make("DataType", DataType::F32));
+} // namespace
+
+TEST_SUITE(NEON)
+TEST_SUITE(ElementwiseDivision)
+
+template <typename T>
+using NEElementwiseDivisionFixture = ElementwiseDivisionValidationFixture<Tensor, Accessor, NEElementwiseDivision, T>;
+
+// *INDENT-OFF*
+// clang-format off
+DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
+ framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Invalid data type combination
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32), // Mismatching shapes
+ }),
+ framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::S32),
+ TensorInfo(TensorShape(48U, 11U, 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(27U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::F32),
+ TensorInfo(TensorShape(48U, 11U, 2U), 1, DataType::F32),
+ })),
+ framework::dataset::make("Expected", { true, true, true, false, false})),
+ input1_info, input2_info, output_info, expected)
+{
+ ARM_COMPUTE_EXPECT(bool(NEElementwiseDivision::validate(&input1_info.clone()->set_is_resizable(false), &input2_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false))) == expected, framework::LogLevel::ERRORS);
+}
+// clang-format on
+// *INDENT-ON*
+
+TEST_SUITE(Float)
+#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
+TEST_SUITE(F16)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<half>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseDivisionFP16Dataset))
+{
+ // Validate output
+ validate(Accessor(_target), _reference, tolerance_fp16, 0.01);
+}
+TEST_SUITE_END() // F16
+#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
+
+TEST_SUITE(F32)
+DATA_TEST_CASE(Configuration, framework::DatasetMode::ALL, framework::dataset::concat(datasets::SmallShapes(), datasets::LargeShapes()),
+ shape)
+{
+ // Create tensors
+ Tensor ref_src1 = create_tensor<Tensor>(shape, DataType::F32);
+ Tensor ref_src2 = create_tensor<Tensor>(shape, DataType::F32);
+ Tensor dst = create_tensor<Tensor>(shape, DataType::F32);
+
+ // Create and Configure function
+ NEElementwiseDivision add;
+ add.configure(&ref_src1, &ref_src2, &dst);
+
+ // Validate valid region
+ const ValidRegion valid_region = shape_to_valid_region(shape);
+ validate(dst.info()->valid_region(), valid_region);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseDivisionFP32Dataset))
+{
+ // Validate output
+ validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLarge, NEElementwiseDivisionFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapes(), ElementwiseDivisionFP32Dataset))
+{
+ // Validate output
+ validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
+}
+
+template <typename T>
+using NEElementwiseDivisionBroadcastFixture = ElementwiseDivisionBroadcastValidationFixture<Tensor, Accessor, NEElementwiseDivision, T>;
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEElementwiseDivisionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapesBroadcast(),
+ ElementwiseDivisionFP32Dataset))
+{
+ // Validate output
+ validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
+}
+
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, NEElementwiseDivisionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(datasets::LargeShapesBroadcast(),
+ ElementwiseDivisionFP32Dataset))
+{
+ // Validate output
+ validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
+}
+TEST_SUITE_END() // F32
+TEST_SUITE_END() // Float
+
+TEST_SUITE_END() // ElementwiseDivision
+TEST_SUITE_END() // NEON
+} // namespace validation
+} // namespace test
+} // namespace arm_compute
diff --git a/tests/validation/fixtures/ElementwiseOperationsFixture.h b/tests/validation/fixtures/ElementwiseOperationsFixture.h
index 8b88f6a..5303084 100644
--- a/tests/validation/fixtures/ElementwiseOperationsFixture.h
+++ b/tests/validation/fixtures/ElementwiseOperationsFixture.h
@@ -325,6 +325,63 @@
qinfo0, qinfo1, qinfo_out);
}
};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseDivisionBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+ template <typename...>
+ void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type)
+ {
+ ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape0, shape1,
+ data_type0, data_type1, output_data_type,
+ QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+ }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseDivisionValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+ template <typename...>
+ void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type)
+ {
+ ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape, shape,
+ data_type0, data_type1, output_data_type,
+ QuantizationInfo(), QuantizationInfo(), QuantizationInfo());
+ }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseDivisionValidationQuantizedFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+ template <typename...>
+ void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type,
+ QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+ {
+ ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape, shape,
+ data_type0, data_type1, output_data_type,
+ qinfo0, qinfo1, qinfo_out);
+ }
+};
+
+template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
+class ElementwiseDivisionQuantizedBroadcastValidationFixture : public ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>
+{
+public:
+ template <typename...>
+ void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type,
+ QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out)
+
+ {
+ ArithmeticOperationsGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(ArithmeticOperation::DIV, shape0, shape1,
+ data_type0, data_type1, output_data_type,
+ qinfo0, qinfo1, qinfo_out);
+ }
+};
+
} // namespace validation
} // namespace test
} // namespace arm_compute
diff --git a/tests/validation/reference/ElementwiseOperations.cpp b/tests/validation/reference/ElementwiseOperations.cpp
index 1f0d42b..6d533ed 100644
--- a/tests/validation/reference/ElementwiseOperations.cpp
+++ b/tests/validation/reference/ElementwiseOperations.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018 ARM Limited.
+ * Copyright (c) 2018-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -74,7 +74,7 @@
}
T result;
- if(op == ArithmeticOperation::ADD || op == ArithmeticOperation::SUB)
+ if(op == ArithmeticOperation::ADD || op == ArithmeticOperation::SUB || op == ArithmeticOperation::DIV)
{
result = (convert_policy == ConvertPolicy::SATURATE) ? saturate_cast<T>(val) : static_cast<T>(val);
}