4 files changed, 79 insertions, 0 deletions
diff --git a/src/backends/neon/workloads/CMakeLists.txt b/src/backends/neon/workloads/CMakeLists.txt
index 8b2ad63f45..b7638208d1 100644
--- a/src/backends/neon/workloads/CMakeLists.txt
+++ b/src/backends/neon/workloads/CMakeLists.txt
@@ -48,6 +48,8 @@ list(APPEND armnnNeonBackendWorkloads_sources
     NeonPermuteWorkload.hpp
     NeonPooling2dWorkload.cpp
     NeonPooling2dWorkload.hpp
+    NeonQuantizeWorkload.cpp
+    NeonQuantizeWorkload.hpp
     NeonReshapeWorkload.cpp
     NeonReshapeWorkload.hpp
     NeonResizeBilinearWorkload.cpp
diff --git a/src/backends/neon/workloads/NeonQuantizeWorkload.cpp b/src/backends/neon/workloads/NeonQuantizeWorkload.cpp
new file mode 100644
index 0000000000..ef24a7f40b
--- /dev/null
+++ b/src/backends/neon/workloads/NeonQuantizeWorkload.cpp
@@ -0,0 +1,50 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include "NeonQuantizeWorkload.hpp"
+#include "NeonWorkloadUtils.hpp"
+
+#include <neon/NeonTensorHandle.hpp>
+#include <aclCommon/ArmComputeTensorUtils.hpp>
+#include <arm_compute/core/Types.h>
+
+#include <boost/polymorphic_pointer_cast.hpp>
+
+namespace armnn
+{
+using namespace armcomputetensorutils;
+
+arm_compute::Status NeonQuantizeWorkloadValidate(const TensorInfo& input, const TensorInfo& output)
+{
+    const arm_compute::TensorInfo neonInputInfo  = armcomputetensorutils::BuildArmComputeTensorInfo(input);
+    const arm_compute::TensorInfo neonOutputInfo = armcomputetensorutils::BuildArmComputeTensorInfo(output);
+
+    return arm_compute::NEQuantizationLayer::validate(&neonInputInfo, &neonOutputInfo);
+}
+
+NeonQuantizeWorkload::NeonQuantizeWorkload(const QuantizeQueueDescriptor& descriptor,
+     const WorkloadInfo& workloadInfo)
+     : BaseWorkload<QuantizeQueueDescriptor>(descriptor, workloadInfo)
+{
+    arm_compute::ITensor& input = boost::polymorphic_pointer_downcast<INeonTensorHandle>(
+                                                                      m_Data.m_Inputs[0])->GetTensor();
+    arm_compute::ITensor& output = boost::polymorphic_pointer_downcast<INeonTensorHandle>(
+                                                                       m_Data.m_Outputs[0])->GetTensor();
+
+    m_Layer.reset(new arm_compute::NEQuantizationLayer());
+    m_Layer->configure(&input, &output);
+    m_Layer->prepare();
+}
+
+void NeonQuantizeWorkload::Execute() const
+{
+    if (m_Layer)
+    {
+        ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonQuantizeWorkload_Execute");
+        m_Layer->run();
+    }
+}
+
+} // namespace armnn
diff --git a/src/backends/neon/workloads/NeonQuantizeWorkload.hpp b/src/backends/neon/workloads/NeonQuantizeWorkload.hpp
new file mode 100644
index 0000000000..7f5a145371
--- /dev/null
+++ b/src/backends/neon/workloads/NeonQuantizeWorkload.hpp
@@ -0,0 +1,26 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#pragma once
+
+#include <backendsCommon/Workload.hpp>
+
+#include <arm_compute/runtime/NEON/functions/NEQuantizationLayer.h>
+
+namespace armnn {
+
+arm_compute::Status NeonQuantizeWorkloadValidate(const TensorInfo& input, const TensorInfo& output);
+
+class NeonQuantizeWorkload : public BaseWorkload<QuantizeQueueDescriptor>
+{
+public:
+    NeonQuantizeWorkload(const QuantizeQueueDescriptor& descriptor, const WorkloadInfo& workloadInfo);
+    void Execute() const override;
+
+private:
+    mutable std::unique_ptr<arm_compute::NEQuantizationLayer> m_Layer;
+};
+
+} // namespace armnn
diff --git a/src/backends/neon/workloads/NeonWorkloads.hpp b/src/backends/neon/workloads/NeonWorkloads.hpp
index dc02aecfd7..8ad70d7095 100644
--- a/src/backends/neon/workloads/NeonWorkloads.hpp
+++ b/src/backends/neon/workloads/NeonWorkloads.hpp
@@ -26,6 +26,7 @@
 #include "NeonPadWorkload.hpp"
 #include "NeonPermuteWorkload.hpp"
 #include "NeonPooling2dWorkload.hpp"
+#include "NeonQuantizeWorkload.hpp"
 #include "NeonReshapeWorkload.hpp"
 #include "NeonResizeBilinearWorkload.hpp"
 #include "NeonSoftmaxFloatWorkload.hpp"