IVGCVSW-3844 Add CL GATHER Workload

Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: I3d44c05acb40566cd4149417fca5bfd260f301e1

10 files changed, 113 insertions, 3 deletions

diff --git a/src/backends/cl/ClLayerSupport.cpp b/src/backends/cl/ClLayerSupport.cpp
index c8d3816a4c..9c0cf61f9c 100644
--- a/src/backends/cl/ClLayerSupport.cpp
+++ b/src/backends/cl/ClLayerSupport.cpp
@@ -34,6 +34,7 @@
 #include "workloads/ClExpWorkload.hpp"
 #include "workloads/ClFloorFloatWorkload.hpp"
 #include "workloads/ClFullyConnectedWorkload.hpp"
+#include "workloads/ClGatherWorkload.hpp"
 #include "workloads/ClInstanceNormalizationWorkload.hpp"
 #include "workloads/ClL2NormalizationFloatWorkload.hpp"
 #include "workloads/ClLstmFloatWorkload.hpp"
@@ -453,6 +454,18 @@ bool ClLayerSupport::IsFullyConnectedSupported(const TensorInfo& input,
                                    descriptor);
 }
 
+bool ClLayerSupport::IsGatherSupported(const TensorInfo& input0,
+                                       const TensorInfo& input1,
+                                       const TensorInfo& output,
+                                       Optional<std::string&> reasonIfUnsupported) const
+{
+    FORWARD_WORKLOAD_VALIDATE_FUNC(ClGatherWorkloadValidate,
+                                   reasonIfUnsupported,
+                                   input0,
+                                   input1,
+                                   output);
+}
+
 bool ClLayerSupport::IsGreaterSupported(const TensorInfo& input0,
                                         const TensorInfo& input1,
                                         const TensorInfo& output,
diff --git a/src/backends/cl/ClLayerSupport.hpp b/src/backends/cl/ClLayerSupport.hpp
index d785f54387..67fd230ea2 100644
--- a/src/backends/cl/ClLayerSupport.hpp
+++ b/src/backends/cl/ClLayerSupport.hpp
@@ -119,6 +119,11 @@ public:
                                    const FullyConnectedDescriptor& descriptor,
                                    Optional<std::string&> reasonIfUnsupported = EmptyOptional()) const override;
 
+    bool IsGatherSupported(const TensorInfo& input0,
+                           const TensorInfo& input1,
+                           const TensorInfo& output,
+                           Optional<std::string&> reasonIfUnsupported) const override;
+
     ARMNN_DEPRECATED_MSG("Use IsComparisonSupported instead")
     bool IsGreaterSupported(const TensorInfo& input0,
                             const TensorInfo& input1,
diff --git a/src/backends/cl/ClWorkloadFactory.cpp b/src/backends/cl/ClWorkloadFactory.cpp
index cabc3466aa..ff66c6bada 100644
--- a/src/backends/cl/ClWorkloadFactory.cpp
+++ b/src/backends/cl/ClWorkloadFactory.cpp
@@ -302,7 +302,7 @@ std::unique_ptr<IWorkload> ClWorkloadFactory::CreateFullyConnected(const FullyCo
 std::unique_ptr<IWorkload> ClWorkloadFactory::CreateGather(const GatherQueueDescriptor& descriptor,
                                                            const WorkloadInfo& info) const
 {
-    return MakeWorkload<NullWorkload, NullWorkload>(descriptor, info);
+    return MakeWorkload<ClGatherWorkload>(descriptor, info);
 }
 
 std::unique_ptr<IWorkload> ClWorkloadFactory::CreateGreater(const GreaterQueueDescriptor& descriptor,
diff --git a/src/backends/cl/backend.mk b/src/backends/cl/backend.mk
index 740aee6d67..721a9ec26e 100644
--- a/src/backends/cl/backend.mk
+++ b/src/backends/cl/backend.mk
@@ -41,6 +41,7 @@ BACKEND_SOURCES := \
         workloads/ClExpWorkload.cpp \
         workloads/ClFloorFloatWorkload.cpp \
         workloads/ClFullyConnectedWorkload.cpp \
+        workloads/ClGatherWorkload.cpp \
         workloads/ClInstanceNormalizationWorkload.cpp \
         workloads/ClL2NormalizationFloatWorkload.cpp \
         workloads/ClLstmFloatWorkload.cpp \
diff --git a/src/backends/cl/test/ClLayerTests.cpp b/src/backends/cl/test/ClLayerTests.cpp
index e6492f612a..eb7dce6930 100644
--- a/src/backends/cl/test/ClLayerTests.cpp
+++ b/src/backends/cl/test/ClLayerTests.cpp
@@ -465,6 +465,12 @@ ARMNN_AUTO_TEST_CASE(DepthToSpaceNhwcInt16_4, DepthToSpaceTest4<DataType::QSymmS
 // Floor
 ARMNN_AUTO_TEST_CASE(SimpleFloor, SimpleFloorTest<DataType::Float32>)
 
+// Gather
+ARMNN_AUTO_TEST_CASE(Gather1dParamsFloat32, Gather1dParamsFloat32Test)
+ARMNN_AUTO_TEST_CASE(Gather1dParamsUint8, Gather1dParamsUint8Test)
+ARMNN_AUTO_TEST_CASE(GatherMultiDimParamsFloat32, GatherMultiDimParamsFloat32Test)
+ARMNN_AUTO_TEST_CASE(GatherMultiDimParamsUint8, GatherMultiDimParamsUint8Test)
+
 // Reshape
 ARMNN_AUTO_TEST_CASE(SimpleReshapeFloat32, SimpleReshapeTest<DataType::Float32>)
 ARMNN_AUTO_TEST_CASE(SimpleReshapeInt8, SimpleReshapeTest<DataType::QAsymmS8>)
diff --git a/src/backends/cl/workloads/CMakeLists.txt b/src/backends/cl/workloads/CMakeLists.txt
index 7d9df07ed9..6baeae0a80 100644
--- a/src/backends/cl/workloads/CMakeLists.txt
+++ b/src/backends/cl/workloads/CMakeLists.txt
@@ -42,6 +42,8 @@ list(APPEND armnnClBackendWorkloads_sources
     ClFloorFloatWorkload.hpp
     ClFullyConnectedWorkload.cpp
     ClFullyConnectedWorkload.hpp
+    ClGatherWorkload.cpp
+    ClGatherWorkload.hpp
     ClInstanceNormalizationWorkload.cpp
     ClInstanceNormalizationWorkload.hpp
     ClL2NormalizationFloatWorkload.cpp
diff --git a/src/backends/cl/workloads/ClConstantWorkload.cpp b/src/backends/cl/workloads/ClConstantWorkload.cpp
index bae7446753..d6b5c57a7e 100644
--- a/src/backends/cl/workloads/ClConstantWorkload.cpp
+++ b/src/backends/cl/workloads/ClConstantWorkload.cpp
@@ -19,14 +19,15 @@ arm_compute::Status ClConstantWorkloadValidate(const TensorInfo& output)
 {
     const arm_compute::TensorInfo neonOutputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(output);
 
-    std::array<arm_compute::DataType,7> supportedTypes = {
+    std::array<arm_compute::DataType,8> supportedTypes = {
             arm_compute::DataType::F16,
             arm_compute::DataType::F32,
             arm_compute::DataType::QASYMM8,
             arm_compute::DataType::QASYMM8_SIGNED,
             arm_compute::DataType::QSYMM16,
             arm_compute::DataType::QSYMM8,
-            arm_compute::DataType::QSYMM8_PER_CHANNEL
+            arm_compute::DataType::QSYMM8_PER_CHANNEL,
+            arm_compute::DataType::S32
     };
     auto it = std::find(begin(supportedTypes), end(supportedTypes), neonOutputInfo.data_type());
 
@@ -95,6 +96,11 @@ void ClConstantWorkload::Execute() const
                 CopyArmComputeClTensorData(output, data.m_LayerOutput->GetConstTensor<int8_t>());
                 break;
             }
+            case arm_compute::DataType::S32:
+            {
+                CopyArmComputeClTensorData(output, data.m_LayerOutput->GetConstTensor<int32_t>());
+                break;
+            }
             default:
             {
                 ARMNN_ASSERT_MSG(false, "Unknown data type");
diff --git a/src/backends/cl/workloads/ClGatherWorkload.cpp b/src/backends/cl/workloads/ClGatherWorkload.cpp
new file mode 100644
index 0000000000..068487039b
--- /dev/null
+++ b/src/backends/cl/workloads/ClGatherWorkload.cpp
@@ -0,0 +1,48 @@
+//
+// Copyright © 2020 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "ClGatherWorkload.hpp"
+#include "ClWorkloadUtils.hpp"
+#include <aclCommon/ArmComputeUtils.hpp>
+#include <cl/ClTensorHandle.hpp>
+
+using namespace armnn::armcomputetensorutils;
+
+namespace armnn
+{
+arm_compute::Status ClGatherWorkloadValidate(const TensorInfo& input,
+                                             const TensorInfo& indices,
+                                             const TensorInfo& output)
+{
+    const arm_compute::TensorInfo aclInput   = BuildArmComputeTensorInfo(input);
+    const arm_compute::TensorInfo aclIndices = BuildArmComputeTensorInfo(indices);
+    const arm_compute::TensorInfo aclOutput  = BuildArmComputeTensorInfo(output);
+
+    int aclAxis = ComputeAclAxis(0, input);
+
+    return arm_compute::CLGather::validate(&aclInput, &aclIndices, &aclOutput, aclAxis);
+}
+
+ClGatherWorkload::ClGatherWorkload(const GatherQueueDescriptor& descriptor,
+                                   const WorkloadInfo& info)
+        : BaseWorkload<GatherQueueDescriptor>(descriptor, info)
+{
+    m_Data.ValidateInputsOutputs("ClGatherWorkload", 1, 1);
+
+    arm_compute::ICLTensor& input    = static_cast<IClTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
+    arm_compute::ICLTensor& indices  = static_cast<IClTensorHandle*>(m_Data.m_Inputs[1])->GetTensor();
+    arm_compute::ICLTensor& output   = static_cast<IClTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();
+
+    int aclAxis = ComputeAclAxis(0, info.m_InputTensorInfos[0]);
+
+    m_Layer.configure(&input, &indices, &output, aclAxis);
+};
+
+void ClGatherWorkload::Execute() const
+{
+    ARMNN_SCOPED_PROFILING_EVENT_CL("ClGatherWorkload_Execute");
+    RunClFunction(m_Layer, CHECK_LOCATION());
+}
+} // namespace armnn
diff --git a/src/backends/cl/workloads/ClGatherWorkload.hpp b/src/backends/cl/workloads/ClGatherWorkload.hpp
new file mode 100644
index 0000000000..5dbeaade59
--- /dev/null
+++ b/src/backends/cl/workloads/ClGatherWorkload.hpp
@@ -0,0 +1,28 @@
+//
+// Copyright © 2020 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include <backendsCommon/Workload.hpp>
+
+#include <arm_compute/runtime/CL/functions/CLGather.h>
+
+namespace armnn
+{
+arm_compute::Status ClGatherWorkloadValidate(const TensorInfo& input,
+                                             const TensorInfo& indices,
+                                             const TensorInfo& output);
+
+class ClGatherWorkload : public BaseWorkload<GatherQueueDescriptor>
+{
+public:
+    ClGatherWorkload(const GatherQueueDescriptor& descriptor, const WorkloadInfo& info);
+    void Execute() const override;
+
+private:
+    mutable arm_compute::CLGather m_Layer;
+};
+
+} // namespace armnn
diff --git a/src/backends/cl/workloads/ClWorkloads.hpp b/src/backends/cl/workloads/ClWorkloads.hpp
index 1ae9a91b88..5c81079ad7 100644
--- a/src/backends/cl/workloads/ClWorkloads.hpp
+++ b/src/backends/cl/workloads/ClWorkloads.hpp
@@ -20,6 +20,7 @@
 #include "ClExpWorkload.hpp"
 #include "ClFloorFloatWorkload.hpp"
 #include "ClFullyConnectedWorkload.hpp"
+#include "ClGatherWorkload.hpp"
 #include "ClInstanceNormalizationWorkload.hpp"
 #include "ClL2NormalizationFloatWorkload.hpp"
 #include "ClLstmFloatWorkload.hpp"