COMPMID-3480: Perform in-place computations in NEArithmeticAdditionKernel

Change-Id: I0089657dd95d7c7b8592984def8e8de1d7e6d085
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3308
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

5 files changed, 125 insertions, 183 deletions

diff --git a/tests/datasets/ShapeDatasets.h b/tests/datasets/ShapeDatasets.h
index 087342d3b8..0ea8091258 100644
--- a/tests/datasets/ShapeDatasets.h
+++ b/tests/datasets/ShapeDatasets.h
@@ -241,6 +241,30 @@ public:
     }
 };
 
+/** Data set containing pairs of small tensor shapes that are broadcast compatible for in-place computations.
+ *  When doing in-place computations, only the second input can be broadcast.
+ */
+class SmallShapesBroadcastInPlace final : public framework::dataset::ZipDataset<ShapeDataset, ShapeDataset>
+{
+public:
+    SmallShapesBroadcastInPlace()
+        : ZipDataset<ShapeDataset, ShapeDataset>(
+              ShapeDataset("Shape0",
+    {
+        TensorShape{ 9U, 9U, 3U, 4U },
+                     TensorShape{ 27U, 13U, 2U, 4U },
+                     TensorShape{ 1U, 16U, 10U, 2U, 128U },
+    }),
+    ShapeDataset("Shape1",
+    {
+        TensorShape{ 9U, 1U, 3U },
+        TensorShape{ 1U },
+        TensorShape{ 1U, 1U, 1U, 1U, 128U },
+    }))
+    {
+    }
+};
+
 /** Data set containing medium tensor shapes. */
 class MediumShapes final : public ShapeDataset
 {
diff --git a/tests/validation/CL/ArithmeticAddition.cpp b/tests/validation/CL/ArithmeticAddition.cpp
index 41415ee481..229dcbc712 100644
--- a/tests/validation/CL/ArithmeticAddition.cpp
+++ b/tests/validation/CL/ArithmeticAddition.cpp
@@ -108,31 +108,9 @@ using CLArithmeticAdditionFixture = ArithmeticAdditionValidationFixture<CLTensor
 
 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)
-{
-    // Create tensors
-    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::U8);
-    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::U8);
-    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::U8);
-
-    // Create and Configure function
-    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<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionU8Dataset),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionU8Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -140,39 +118,17 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<uint8_t>, framework
 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 })))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  framework::dataset::make("InPlace", { false })))
 {
     // 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(RunLarge, CLArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -185,29 +141,6 @@ using CLArithmeticAdditionQuantizedFixture = ArithmeticAdditionValidationQuantiz
 
 TEST_SUITE(Quantized)
 TEST_SUITE(QASYMM8)
-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, DataType::QASYMM8);
-    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QASYMM8);
-    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8);
-
-    // 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, CLArithmeticAdditionQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ArithmeticAdditionQASYMM8Dataset),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
@@ -220,29 +153,6 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionQuantizedFixture<uint8_t>,
 }
 TEST_SUITE_END() // QASYMM8
 TEST_SUITE(QASYMM8_SIGNED)
-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, DataType::QASYMM8_SIGNED);
-    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::QASYMM8_SIGNED);
-    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::QASYMM8_SIGNED);
-
-    // 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, CLArithmeticAdditionQuantizedFixture<int8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ArithmeticAdditionQASYMM8SignedDataset),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
@@ -255,29 +165,6 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLArithmeticAdditionQuantizedFixture<int8_t>, f
 }
 TEST_SUITE_END() // QASYMM8_SIGNED
 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, 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;
-    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, CLArithmeticAdditionQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ArithmeticAdditionQSYMM16Dataset),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
@@ -313,29 +200,6 @@ FIXTURE_DATA_TEST_CASE(RunWithActivation, CLArithmeticAdditionFloatFixture<half>
 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 })),
-               shape, policy)
-{
-    // Create tensors
-    CLTensor ref_src1 = create_tensor<CLTensor>(shape, DataType::F32);
-    CLTensor ref_src2 = create_tensor<CLTensor>(shape, DataType::F32);
-    CLTensor dst      = create_tensor<CLTensor>(shape, DataType::F32);
-
-    // Create and Configure function
-    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, CLArithmeticAdditionFloatFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP32Dataset),
                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
                                                                                                                      EmptyActivationFunctionsDataset))
diff --git a/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp b/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
index 82946fa209..336870ac8a 100755..100644
--- a/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
+++ b/tests/validation/GLES_COMPUTE/ArithmeticAddition.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2018 ARM Limited.
+ * Copyright (c) 2017-2020 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -56,8 +56,9 @@ using GCArithmeticAdditionFixture = ArithmeticAdditionValidationFixture<GCTensor
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
-FIXTURE_DATA_TEST_CASE(RunSmall, GCArithmeticAdditionFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP16Dataset),
-                                                                                                         framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, GCArithmeticAdditionFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP16Dataset),
+                                                                                                                 framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                         framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(GCAccessor(_target), _reference);
diff --git a/tests/validation/NEON/ArithmeticAddition.cpp b/tests/validation/NEON/ArithmeticAddition.cpp
index 72993172fd..19ebfedcb5 100644
--- a/tests/validation/NEON/ArithmeticAddition.cpp
+++ b/tests/validation/NEON/ArithmeticAddition.cpp
@@ -104,13 +104,33 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(
                                               ConvertPolicy::WRAP);
     ARM_COMPUTE_EXPECT(bool(s) == expected, framework::LogLevel::ERRORS);
 }
+
+DATA_TEST_CASE(ValidateInPlace, framework::DatasetMode::ALL, zip(zip(
+               framework::dataset::make("Input1Info", { TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::U8),
+                                                        TensorInfo(TensorShape(1U, 13U, 2U), 1, DataType::U8), // Unsupported broadcast on input1 with in-place computation
+                                                      }),
+               framework::dataset::make("Input2Info",{ TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                       TensorInfo(TensorShape(1U, 13U, 2U), 1, DataType::U8),
+                                                       TensorInfo(TensorShape(32U, 13U, 2U), 1, DataType::U8),
+                                                     })),
+               framework::dataset::make("Expected", { true, true, false })),
+               input1_info, input2_info, expected)
+{
+    Status s = NEArithmeticAddition::validate(&input1_info.clone()->set_is_resizable(false),
+                                              &input2_info.clone()->set_is_resizable(false),
+                                              nullptr,
+                                              ConvertPolicy::WRAP);
+    ARM_COMPUTE_EXPECT(bool(s) == expected, framework::LogLevel::ERRORS);
+}
 // clang-format on
 // *INDENT-ON*
 
 TEST_SUITE(Integer)
 TEST_SUITE(U8)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionU8Dataset),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionU8Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  framework::dataset::make("InPlace", { false, true })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -118,15 +138,17 @@ FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<uint8_t>, framework
 TEST_SUITE_END() // U8
 
 TEST_SUITE(S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionS16Dataset),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticAdditionS16Dataset),
-                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), ArithmeticAdditionS16Dataset),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -134,8 +156,9 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<int16_t>, framework
 TEST_SUITE_END() // S16
 
 TEST_SUITE(S32)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int32_t>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ArithmeticAdditionS32Dataset),
-                                                                                                            framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int32_t>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionS32Dataset),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                            framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -146,8 +169,9 @@ TEST_SUITE_END() // Integer
 TEST_SUITE(Float)
 #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 TEST_SUITE(F16)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP16Dataset),
-                                                                                                         framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP16Dataset),
+                                                                                                                 framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                         framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -156,15 +180,17 @@ TEST_SUITE_END() // F16
 #endif           /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
 
 TEST_SUITE(F32)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP32Dataset),
-                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), ArithmeticAdditionFP32Dataset),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), ArithmeticAdditionFP32Dataset),
-                                                                                                              framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), ArithmeticAdditionFP32Dataset),
+                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                              framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -173,17 +199,28 @@ FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<float>, framework::
 template <typename T>
 using NEArithmeticAdditionBroadcastFixture = ArithmeticAdditionBroadcastValidationFixture<Tensor, Accessor, NEArithmeticAddition, T>;
 
-FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticAdditionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapesBroadcast(),
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticAdditionBroadcastFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapesBroadcast(),
+                       ArithmeticAdditionFP32Dataset),
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       framework::dataset::make("InPlace", { false })))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcastInPlace, NEArithmeticAdditionBroadcastFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapesBroadcastInPlace(),
                        ArithmeticAdditionFP32Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       framework::dataset::make("InPlace", { true })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, NEArithmeticAdditionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapesBroadcast(),
+FIXTURE_DATA_TEST_CASE(RunLargeBroadcast, NEArithmeticAdditionBroadcastFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapesBroadcast(),
                        ArithmeticAdditionFP32Dataset),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       framework::dataset::make("InPlace", { false })))
 {
     // Validate output
     validate(Accessor(_target), _reference);
diff --git a/tests/validation/fixtures/ArithmeticOperationsFixture.h b/tests/validation/fixtures/ArithmeticOperationsFixture.h
index 4a6b0bd3f3..1019e60233 100644
--- a/tests/validation/fixtures/ArithmeticOperationsFixture.h
+++ b/tests/validation/fixtures/ArithmeticOperationsFixture.h
@@ -48,8 +48,10 @@ public:
     template <typename...>
     void setup(reference::ArithmeticOperation op, const TensorShape &shape0, const TensorShape &shape1,
                DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy,
-               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out, ActivationLayerInfo act_info)
+               QuantizationInfo qinfo0, QuantizationInfo qinfo1, QuantizationInfo qinfo_out, ActivationLayerInfo act_info,
+               bool in_place)
     {
+        _in_place  = in_place;
         _op        = op;
         _act_info  = act_info;
         _target    = compute_target(shape0, shape1, data_type0, data_type1, output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out);
@@ -71,9 +73,11 @@ protected:
         TensorType ref_src2 = create_tensor<TensorType>(shape1, data_type1, 1, qinfo1);
         TensorType dst      = create_tensor<TensorType>(TensorShape::broadcast_shape(shape0, shape1), output_data_type, 1, qinfo_out);
 
+        TensorType *dst_ptr = _in_place ? nullptr : &dst;
+
         // Create and configure function
         FunctionType arith_op;
-        arith_op.configure(&ref_src1, &ref_src2, &dst, convert_policy, _act_info);
+        arith_op.configure(&ref_src1, &ref_src2, dst_ptr, convert_policy, _act_info);
 
         ARM_COMPUTE_EXPECT(ref_src1.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(ref_src2.info()->is_resizable(), framework::LogLevel::ERRORS);
@@ -82,11 +86,15 @@ protected:
         // Allocate tensors
         ref_src1.allocator()->allocate();
         ref_src2.allocator()->allocate();
-        dst.allocator()->allocate();
 
         ARM_COMPUTE_EXPECT(!ref_src1.info()->is_resizable(), framework::LogLevel::ERRORS);
         ARM_COMPUTE_EXPECT(!ref_src2.info()->is_resizable(), framework::LogLevel::ERRORS);
-        ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+
+        if(!_in_place)
+        {
+            dst.allocator()->allocate();
+            ARM_COMPUTE_EXPECT(!dst.info()->is_resizable(), framework::LogLevel::ERRORS);
+        }
 
         // Fill tensors
         fill(AccessorType(ref_src1), 0);
@@ -95,7 +103,14 @@ protected:
         // Compute function
         arith_op.run();
 
-        return dst;
+        if(_in_place)
+        {
+            return ref_src1;
+        }
+        else
+        {
+            return dst;
+        }
     }
 
     SimpleTensor<T> compute_reference(const TensorShape &shape0, const TensorShape &shape1,
@@ -119,6 +134,7 @@ protected:
     SimpleTensor<T>                _reference{};
     reference::ArithmeticOperation _op{ reference::ArithmeticOperation::ADD };
     ActivationLayerInfo            _act_info{};
+    bool                           _in_place{};
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
@@ -126,10 +142,10 @@ class ArithmeticAdditionBroadcastValidationFixture : public ArithmeticOperationG
 {
 public:
     template <typename...>
-    void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
+    void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy, bool in_place)
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape0, shape1, data_type0, data_type1,
-                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo());
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo(), in_place);
     }
 };
 
@@ -138,10 +154,10 @@ class ArithmeticAdditionValidationFixture : public ArithmeticOperationGenericFix
 {
 public:
     template <typename...>
-    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy)
+    void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy, bool in_place)
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape, shape, data_type0, data_type1,
-                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo());
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo(), in_place);
     }
 };
 
@@ -153,7 +169,7 @@ public:
     void setup(const TensorShape &shape0, const TensorShape &shape1, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy, ActivationLayerInfo act_info)
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape0, shape1, data_type0, data_type1,
-                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info);
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info, false);
     }
 };
 
@@ -165,7 +181,7 @@ public:
     void setup(const TensorShape &shape, DataType data_type0, DataType data_type1, DataType output_data_type, ConvertPolicy convert_policy, ActivationLayerInfo act_info)
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape, shape, data_type0, data_type1,
-                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info);
+                                                                                            output_data_type, convert_policy, QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info, false);
     }
 };
 
@@ -179,7 +195,7 @@ public:
 
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::ADD, shape, shape, data_type0, data_type1,
-                                                                                            output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out, ActivationLayerInfo());
+                                                                                            output_data_type, convert_policy, qinfo0, qinfo1, qinfo_out, ActivationLayerInfo(), false);
     }
 };
 
@@ -192,7 +208,7 @@ public:
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape0, shape1,
                                                                                             data_type0, data_type1, output_data_type, convert_policy,
-                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo());
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo(), false);
     }
 };
 
@@ -205,7 +221,7 @@ public:
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape0, shape1,
                                                                                             data_type0, data_type1, output_data_type, convert_policy,
-                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info);
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info, false);
     }
 };
 
@@ -218,7 +234,7 @@ public:
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape, shape,
                                                                                             data_type0, data_type1, output_data_type, convert_policy,
-                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo());
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), ActivationLayerInfo(), false);
     }
 };
 
@@ -231,7 +247,7 @@ public:
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape, shape,
                                                                                             data_type0, data_type1, output_data_type, convert_policy,
-                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info);
+                                                                                            QuantizationInfo(), QuantizationInfo(), QuantizationInfo(), act_info, false);
     }
 };
 
@@ -246,7 +262,7 @@ public:
     {
         ArithmeticOperationGenericFixture<TensorType, AccessorType, FunctionType, T>::setup(reference::ArithmeticOperation::SUB, shape, shape,
                                                                                             data_type0, data_type1, output_data_type,
-                                                                                            convert_policy, qinfo0, qinfo1, qinfo_out, ActivationLayerInfo());
+                                                                                            convert_policy, qinfo0, qinfo1, qinfo_out, ActivationLayerInfo(), false);
     }
 };
 } // namespace validation