IVGCVSW-5775 'Add Async Support to ExecuteNetwork'

* Enabled async mode with '-n, concurrent' and 'simultaneous-iterations'
  in ExecuteNetwork
* Number of input files provided should be equal to number of input files
  provided multiply by number of simultaneous iterations divided by comma

!armnn:5443

Signed-off-by: Sadik Armagan <sadik.armagan@arm.com>
Change-Id: Ibeb318010430bf4ae61a02b18b1bf88f3657774c

3 files changed, 12 insertions, 5 deletions

diff --git a/src/armnn/WorkingMemHandle.cpp b/src/armnn/WorkingMemHandle.cpp
index 0cbef82e83..b54c5baddd 100644
--- a/src/armnn/WorkingMemHandle.cpp
+++ b/src/armnn/WorkingMemHandle.cpp
@@ -26,7 +26,8 @@ WorkingMemHandle::WorkingMemHandle(
     m_MemoryManagers(memoryManagers),
     m_OwnedTensorHandles(std::move(ownedTensorHandles)),
     m_IsAllocated(false),
-    m_Mutex()
+    m_Mutex(),
+    m_InferenceId(profiling::ProfilingService::GetNextGuid())
 {
 }
 
diff --git a/src/armnn/WorkingMemHandle.hpp b/src/armnn/WorkingMemHandle.hpp
index 92b0acaec3..5ccb2b2342 100644
--- a/src/armnn/WorkingMemHandle.hpp
+++ b/src/armnn/WorkingMemHandle.hpp
@@ -38,6 +38,11 @@ public:
         return m_NetworkId;
     }
 
+    profiling::ProfilingGuid GetInferenceId() override
+    {
+        return m_InferenceId;
+    }
+
     /// Allocate the backing memory required for execution. If this is not called, then allocation will be
     /// deferred to execution time. The mutex must be locked.
     void Allocate() override;
@@ -87,6 +92,7 @@ private:
 
     bool m_IsAllocated;
     std::mutex m_Mutex;
+    profiling::ProfilingGuid m_InferenceId;
 };
 
 } // end experimental namespace
diff --git a/src/backends/reference/workloads/RefDetectionPostProcessWorkload.cpp b/src/backends/reference/workloads/RefDetectionPostProcessWorkload.cpp
index 25c326ad37..6784e21585 100644
--- a/src/backends/reference/workloads/RefDetectionPostProcessWorkload.cpp
+++ b/src/backends/reference/workloads/RefDetectionPostProcessWorkload.cpp
@@ -46,10 +46,10 @@ void RefDetectionPostProcessWorkload::Execute(std::vector<ITensorHandle*> inputs
     auto scores       = MakeDecoder<float>(scoresInfo, inputs[1]->Map());
     auto anchors      = MakeDecoder<float>(anchorsInfo, m_Anchors->Map(false));
 
-    float* detectionBoxes   = GetOutputTensorData<float>(0, m_Data);
-    float* detectionClasses = GetOutputTensorData<float>(1, m_Data);
-    float* detectionScores  = GetOutputTensorData<float>(2, m_Data);
-    float* numDetections    = GetOutputTensorData<float>(3, m_Data);
+    float* detectionBoxes   = reinterpret_cast<float*>(outputs[0]->Map());
+    float* detectionClasses = reinterpret_cast<float*>(outputs[1]->Map());
+    float* detectionScores  = reinterpret_cast<float*>(outputs[2]->Map());
+    float* numDetections    = reinterpret_cast<float*>(outputs[3]->Map());
 
     DetectionPostProcess(boxEncodingsInfo, scoresInfo, anchorsInfo,
                          detectionBoxesInfo, detectionClassesInfo,