blob: 79a0b267e90827155eb1c3246f4119993625b292 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
|
//
// Copyright © 2017 Arm Ltd. All rights reserved.
// See LICENSE file in the project root for full license information.
//
#pragma once
#include "armnn/Tensor.hpp"
#include "armnn/Types.hpp"
#include "Network.hpp"
#include "LayerFwd.hpp"
#include "backends/RefWorkloadFactory.hpp"
#include "backends/NeonWorkloadFactory.hpp"
#include "backends/ClWorkloadFactory.hpp"
#include "backends/Workload.hpp"
#include "backends/WorkloadFactory.hpp"
namespace cl
{
class Context;
class CommandQueue;
class Device;
}
namespace armnn
{
class LoadedNetwork
{
public:
TensorInfo GetInputTensorInfo(LayerBindingId layerId) const;
TensorInfo GetOutputTensorInfo(LayerBindingId layerId) const;
Status EnqueueWorkload(const InputTensors& inputTensors, const OutputTensors& outputTensors);
static std::unique_ptr<LoadedNetwork> MakeLoadedNetwork(std::unique_ptr<OptimizedNetwork> net,
bool useCpuRefAsFallback);
private:
LoadedNetwork(std::unique_ptr<OptimizedNetwork> net, bool useCpuRefAsFallback);
void EnqueueInput(const BindableLayer& layer, ITensorHandle* tensorHandle, const TensorInfo& tensorInfo);
void EnqueueOutput(const BindableLayer& layer, ITensorHandle* tensorHandle, const TensorInfo& tensorInfo);
bool Execute();
void TidyWorkloadQueue(size_t numInputs, size_t numOutputs);
const IWorkloadFactory& GetWorkloadFactory(const Layer& layer) const;
RefWorkloadFactory m_CpuRef;
NeonWorkloadFactory m_CpuAcc;
ClWorkloadFactory m_GpuAcc;
std::unique_ptr<OptimizedNetwork> m_OptimizedNetwork;
std::vector< std::unique_ptr<IWorkload> > m_WorkloadQueue;
};
}
|
