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, 30 insertions, 18 deletions

diff --git a/tests/validation/CL/ElementwiseSquaredDiff.cpp b/tests/validation/CL/ElementwiseSquaredDiff.cpp
index 4c732b0885..ee0279df33 100644
--- a/tests/validation/CL/ElementwiseSquaredDiff.cpp
+++ b/tests/validation/CL/ElementwiseSquaredDiff.cpp
@@ -70,6 +70,8 @@ const auto ActivationFunctionsDataset = framework::dataset::make("ActivationInfo
     ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::BOUNDED_RELU, 0.75f, 0.25f),
     ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::LOGISTIC, 0.75f, 0.25f)
 });
+const auto InPlaceDataSet    = framework::dataset::make("InPlace", { false, true });
+const auto OutOfPlaceDataSet = framework::dataset::make("InPlace", { false });
 } // namespace
 
 TEST_SUITE(CL)
@@ -103,7 +105,8 @@ using CLElementwiseSquaredDiffFixture = ElementwiseSquaredDiffValidationFixture<
 
 TEST_SUITE(Integer)
 TEST_SUITE(U8)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(datasets::SmallShapes(), ElementwiseSquaredDiffU8Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffU8Dataset),
+                                                                                                                      OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -111,7 +114,8 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<uint8_t>, frame
 TEST_SUITE_END()
 
 TEST_SUITE(S16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<int16_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseSquaredDiffS16Dataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFixture<int16_t>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffS16Dataset),
+                                                                                                                OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference);
@@ -124,22 +128,24 @@ using CLElementwiseSquaredDiffQuantizedFixture = ElementwiseSquaredDiffValidatio
 
 TEST_SUITE(Quantized)
 TEST_SUITE(QASYMM8)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ElementwiseSquaredDiffQASYMM8Dataset),
                        framework::dataset::make("Src0QInfo", { QuantizationInfo(5.f / 255.f, 20) })),
                        framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 255.f, 10) })),
-                       framework::dataset::make("OutQInfo", { QuantizationInfo(1.f / 255.f, 5) })))
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(1.f / 255.f, 5) })),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32, 0.01);
 }
 TEST_SUITE_END()
 TEST_SUITE(QSYMM16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(datasets::SmallShapes(),
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffQuantizedFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(combine(combine(datasets::SmallShapes(),
                        ElementwiseSquaredDiffQSYMM16Dataset),
                        framework::dataset::make("Src0QInfo", { QuantizationInfo(1.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
                        framework::dataset::make("Src1QInfo", { QuantizationInfo(2.f / 32768.f, 0), QuantizationInfo(5.f / 32768.f, 0) })),
-                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })))
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(5.f / 32768.f, 0) })),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_qsymm16);
@@ -152,14 +158,16 @@ using CLElementwiseSquaredDiffFloatFixture = ElementwiseSquaredDiffValidationFlo
 
 TEST_SUITE(Float)
 TEST_SUITE(FP16)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFloatFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP16Dataset),
-                                                                                                                  EmptyActivationFunctionsDataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFloatFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP16Dataset),
+                                                                                                                  EmptyActivationFunctionsDataset),
+                                                                                                                  OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
 }
-FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<half>, framework::DatasetMode::ALL, combine(combine(datasets::TinyShapes(), ElementwiseSquaredDiffFP16Dataset),
-                       ActivationFunctionsDataset))
+FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::TinyShapes(), ElementwiseSquaredDiffFP16Dataset),
+                       ActivationFunctionsDataset),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp16, 0.01);
@@ -167,14 +175,16 @@ FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<h
 TEST_SUITE_END()
 
 TEST_SUITE(FP32)
-FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP32Dataset),
-                                                                                                                   EmptyActivationFunctionsDataset))
+FIXTURE_DATA_TEST_CASE(RunSmall, CLElementwiseSquaredDiffFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), ElementwiseSquaredDiffFP32Dataset),
+                                                                                                                   EmptyActivationFunctionsDataset),
+                                                                                                                   OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
 }
-FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::TinyShapes(), ElementwiseSquaredDiffFP32Dataset),
-                       ActivationFunctionsDataset))
+FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::TinyShapes(), ElementwiseSquaredDiffFP32Dataset),
+                       ActivationFunctionsDataset),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
@@ -182,16 +192,18 @@ FIXTURE_DATA_TEST_CASE(RunWithActivation, CLElementwiseSquaredDiffFloatFixture<f
 template <typename T>
 using CLElementwiseSquaredDiffBroadcastFloatFixture = ElementwiseSquaredDiffBroadcastValidationFloatFixture<CLTensor, CLAccessor, CLElementwiseSquaredDiff, T>;
 
-FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseSquaredDiffBroadcastFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::SmallShapesBroadcast(),
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, CLElementwiseSquaredDiffBroadcastFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapesBroadcast(),
                        ElementwiseSquaredDiffFP32Dataset),
-                       EmptyActivationFunctionsDataset))
+                       EmptyActivationFunctionsDataset),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);
 }
-FIXTURE_DATA_TEST_CASE(RunWithActivationBroadcast, CLElementwiseSquaredDiffBroadcastFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(datasets::TinyShapesBroadcast(),
+FIXTURE_DATA_TEST_CASE(RunWithActivationBroadcast, CLElementwiseSquaredDiffBroadcastFloatFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(datasets::TinyShapesBroadcast(),
                        ElementwiseSquaredDiffFP32Dataset),
-                       ActivationFunctionsDataset))
+                       ActivationFunctionsDataset),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(CLAccessor(_target), _reference, tolerance_fp32);