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, 56 insertions, 35 deletions

diff --git a/tests/validation/NEON/ArithmeticAddition.cpp b/tests/validation/NEON/ArithmeticAddition.cpp
index 213dbc1f5e..f3e4dfc6e5 100644
--- a/tests/validation/NEON/ArithmeticAddition.cpp
+++ b/tests/validation/NEON/ArithmeticAddition.cpp
@@ -48,6 +48,8 @@ constexpr AbsoluteTolerance<float> tolerance_quant(1); /**< Tolerance value for
 #else                                                  // !defined(__aarch64__) || defined(ENABLE_SVE)
 constexpr AbsoluteTolerance<float> tolerance_quant(0);
 #endif                                                 // !defined(__aarch64__) || defined(ENABLE_SVE)
+const auto InPlaceDataSet    = framework::dataset::make("InPlace", { false, true });
+const auto OutOfPlaceDataSet = framework::dataset::make("InPlace", { false });
 } // namespace
 
 TEST_SUITE(NEON)
@@ -104,8 +106,10 @@ TEST_CASE(NoPaddingAdded, framework::DatasetMode::PRECOMMIT)
 
 TEST_SUITE(Integer)
 TEST_SUITE(U8)
-FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::U8)),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<uint8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType",
+                                                                                                                  DataType::U8)),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -113,15 +117,19 @@ 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(), framework::dataset::make("DataType", DataType::S16)),
-                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType",
+                                                                                                                  DataType::S16)),
+                                                                                                                  framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                  OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), framework::dataset::make("DataType", DataType::S16)),
-                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), framework::dataset::make("DataType",
+                                                                                                                        DataType::S16)),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -129,8 +137,10 @@ 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(), framework::dataset::make("DataType", DataType::S32)),
-                                                                                                            framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<int32_t>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType",
+                                                                                                                    DataType::S32)),
+                                                                                                                    framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                            OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -141,8 +151,9 @@ 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(), framework::dataset::make("DataType", DataType::F16)),
-                                                                                                         framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::F16)),
+                                                                                                                 framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                         OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -151,15 +162,19 @@ 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(), framework::dataset::make("DataType", DataType::F32)),
-                                                                                                                framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunSmall, NEArithmeticAdditionFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType",
+                                                                                                                        DataType::F32)),
+                                                                                                                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                                OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
 }
 
-FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeShapes(), framework::dataset::make("DataType", DataType::F32)),
-                                                                                                              framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+FIXTURE_DATA_TEST_CASE(RunLarge, NEArithmeticAdditionFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeShapes(), framework::dataset::make("DataType",
+                                                                                                                      DataType::F32)),
+                                                                                                                      framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                                                                                                              OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -168,17 +183,19 @@ 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(),
                        framework::dataset::make("DataType", DataType::F32)),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       OutOfPlaceDataSet))
 {
     // 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(),
                        framework::dataset::make("DataType", DataType::F32)),
-                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })))
+                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE, ConvertPolicy::WRAP })),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference);
@@ -197,11 +214,12 @@ TEST_SUITE(QASYMM8)
 FIXTURE_DATA_TEST_CASE(RunSmall,
                        NEArithmeticAdditionQuantizedFixture<uint8_t>,
                        framework::DatasetMode::PRECOMMIT,
-                       combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QASYMM8)),
-                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
-                                               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) })))
+                       combine(combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QASYMM8)),
+                                                               framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                                                       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) })),
+                               OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_quant);
@@ -212,22 +230,24 @@ TEST_SUITE(QASYMM8_SIGNED)
 FIXTURE_DATA_TEST_CASE(RunSmall,
                        NEArithmeticAdditionQuantizedFixture<int8_t>,
                        framework::DatasetMode::ALL,
-                       combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QASYMM8_SIGNED)),
-                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
-                                               framework::dataset::make("Src0QInfo", { QuantizationInfo(0.5f, 20) })),
-                                       framework::dataset::make("Src1QInfo", { QuantizationInfo(0.5f, 10) })),
-                               framework::dataset::make("OutQInfo", { QuantizationInfo(0.5f, 5) })))
+                       combine(combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QASYMM8_SIGNED)),
+                                                               framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                                                       framework::dataset::make("Src0QInfo", { QuantizationInfo(0.5f, 20) })),
+                                               framework::dataset::make("Src1QInfo", { QuantizationInfo(0.5f, 10) })),
+                                       framework::dataset::make("OutQInfo", { QuantizationInfo(0.5f, 5) })),
+                               OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_quant);
 }
 
-FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticAdditionQuantizedBroadcastFixture<int8_t>, framework::DatasetMode::ALL, combine(combine(combine(combine(combine(
+FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEArithmeticAdditionQuantizedBroadcastFixture<int8_t>, framework::DatasetMode::ALL, combine(combine(combine(combine(combine(combine(
                            datasets::SmallShapesBroadcast(), framework::dataset::make("DataType", DataType::QASYMM8_SIGNED)),
                        framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
                        framework::dataset::make("Src0QInfo", { QuantizationInfo(0.5f, 20) })),
                        framework::dataset::make("Src1QInfo", { QuantizationInfo(0.5f, 10) })),
-                       framework::dataset::make("OutQInfo", { QuantizationInfo(0.5f, 5) })))
+                       framework::dataset::make("OutQInfo", { QuantizationInfo(0.5f, 5) })),
+                       OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_quant);
@@ -238,11 +258,12 @@ TEST_SUITE(QSYMM16)
 FIXTURE_DATA_TEST_CASE(RunSmall,
                        NEArithmeticAdditionQuantizedFixture<int16_t>,
                        framework::DatasetMode::PRECOMMIT,
-                       combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QSYMM16)),
-                                                       framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
-                                               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) })))
+                       combine(combine(combine(combine(combine(combine(datasets::SmallShapes(), framework::dataset::make("DataType", DataType::QSYMM16)),
+                                                               framework::dataset::make("ConvertPolicy", { ConvertPolicy::SATURATE })),
+                                                       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) })),
+                               OutOfPlaceDataSet))
 {
     // Validate output
     validate(Accessor(_target), _reference, tolerance_quant);