Add in-place calculation support for CL elementwise arithmetic kernels

- Add in-place calculation support in ClArithmeticKernel,  ClSaturatedArithmeticKernel and ClMulKernel
- Add in-place test cases

Resolves: COMPMID-4431

Signed-off-by: Sheri Zhang <sheri.zhang@arm.com>
Change-Id: Id484bdb76b74478a33fedb471ae0c7f799c599f6
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5885
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 23 insertions, 8 deletions

diff --git a/tests/validation/NEON/ElementwiseDivision.cpp b/tests/validation/NEON/ElementwiseDivision.cpp
index 8abccb2ed6..5f0224c91d 100644
--- a/tests/validation/NEON/ElementwiseDivision.cpp
+++ b/tests/validation/NEON/ElementwiseDivision.cpp
@@ -56,6 +56,8 @@ const auto              ElementwiseDivisionFP16Dataset = combine(combine(framewo
 #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));
+const auto InPlaceDataSet    = framework::dataset::make("InPlace", { false, true });
+const auto OutOfPlaceDataSet = framework::dataset::make("InPlace", { false });
 } // namespace
 
 TEST_SUITE(NEON)
@@ -105,14 +107,16 @@ using CpuElementwiseDivisionBroadcastDynamicShapeFixture = ArithmeticDivisionBro
 
 TEST_SUITE(F32)
 
-FIXTURE_DATA_TEST_CASE(RunSmall, CpuElementwiseDivisionDynamicShapeFixture<float>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseDivisionFP32Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CpuElementwiseDivisionDynamicShapeFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseDivisionFP32Dataset),
+                                                                                                                        InPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CpuElementwiseDivisionBroadcastDynamicShapeFixture<float>, framework::DatasetMode::ALL, combine(datasets::SmallShapesBroadcast(),
-                       ElementwiseDivisionFP32Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CpuElementwiseDivisionBroadcastDynamicShapeFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapesBroadcast(),
+                       ElementwiseDivisionFP32Dataset),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
@@ -124,7 +128,8 @@ TEST_SUITE_END() // DynamicShape
 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))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseDivisionFP16Dataset),
+                                                                                                          InPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_fp16, 0.01);
@@ -133,7 +138,8 @@ TEST_SUITE_END() // F16
 #endif           /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
 
 TEST_SUITE(F32)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<float>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseDivisionFP32Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseDivisionFP32Dataset),
+                                                                                                           InPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
@@ -142,8 +148,16 @@ FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<float>, framework:
 template <typename T>
 using NEElementwiseDivisionBroadcastFixture = ArithmeticDivisionBroadcastValidationFixture<Tensor, Accessor, NEElementwiseDivision, T>;
 
-FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEElementwiseDivisionBroadcastFixture<float>, framework::DatasetMode::ALL, combine(datasets::SmallShapesBroadcast(),
-                       ElementwiseDivisionFP32Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEElementwiseDivisionBroadcastFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapesBroadcast(),
+                       ElementwiseDivisionFP32Dataset),
+                       OutOfPlaceDataSet))
+{
+    // Validate output
+    validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
+}
+FIXTURE_DATA_TEST_CASE(RunTinyBroadcastInPlace, NEElementwiseDivisionBroadcastFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::TinyShapesBroadcastInplace(),
+                       ElementwiseDivisionFP32Dataset),
+                       InPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_fp32, 0.01);
@@ -153,7 +167,8 @@ TEST_SUITE_END() // Float
 
 TEST_SUITE(Integer)
 TEST_SUITE(S32)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<int32_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseDivisionS32Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<int32_t>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseDivisionS32Dataset),
+                                                                                                             InPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_zero_s32);