From fd627ffaec8fd8801d980b4c91ee7c0607ab6aaf Mon Sep 17 00:00:00 2001 From: Jan Eilers Date: Thu, 25 Feb 2021 17:44:00 +0000 Subject: IVGCVSW-5687 Update Doxygen Docu * Update Doxygen Documentation for 21.02 release Signed-off-by: Jan Eilers Change-Id: I9ed2f9caab038836ea99d7b378d7899fe431a4e5 --- 21.02/_cl_tensor_handle_8hpp_source.xhtml | 161 ++++++++++++++++++++++++++++++ 1 file changed, 161 insertions(+) create mode 100644 21.02/_cl_tensor_handle_8hpp_source.xhtml (limited to '21.02/_cl_tensor_handle_8hpp_source.xhtml') diff --git a/21.02/_cl_tensor_handle_8hpp_source.xhtml b/21.02/_cl_tensor_handle_8hpp_source.xhtml new file mode 100644 index 0000000000..2b1ffb2b96 --- /dev/null +++ b/21.02/_cl_tensor_handle_8hpp_source.xhtml @@ -0,0 +1,161 @@ + + + + + + + + + + + + + +ArmNN: src/backends/cl/ClTensorHandle.hpp Source File + + + + + + + + + + + + + + + + +
+
+ + + + ArmNN + + + +
+
+  21.02 +
+
+
+ + + + + + + +
+
+ +
+
+
+ +
+ +
+
+ + +
+ +
+ +
+
+
ClTensorHandle.hpp
+
+
+Go to the documentation of this file.
1 //
2 // Copyright © 2017 Arm Ltd. All rights reserved.
3 // SPDX-License-Identifier: MIT
4 //
5 #pragma once
6 
7 #include <aclCommon/ArmComputeTensorHandle.hpp>
8 #include <aclCommon/ArmComputeTensorUtils.hpp>
9 
10 #include <Half.hpp>
11 
12 #include <armnn/utility/PolymorphicDowncast.hpp>
13 
14 #include <arm_compute/runtime/CL/CLTensor.h>
15 #include <arm_compute/runtime/CL/CLSubTensor.h>
16 #include <arm_compute/runtime/IMemoryGroup.h>
17 #include <arm_compute/runtime/MemoryGroup.h>
18 #include <arm_compute/core/TensorShape.h>
19 #include <arm_compute/core/Coordinates.h>
20 
21 namespace armnn
22 {
23 
24 
25 class IClTensorHandle : public IAclTensorHandle
26 {
27 public:
28  virtual arm_compute::ICLTensor& GetTensor() = 0;
29  virtual arm_compute::ICLTensor const& GetTensor() const = 0;
30  virtual arm_compute::DataType GetDataType() const = 0;
31  virtual void SetMemoryGroup(const std::shared_ptr<arm_compute::IMemoryGroup>& memoryGroup) = 0;
32 };
33 
34 class ClTensorHandle : public IClTensorHandle
35 {
36 public:
37  ClTensorHandle(const TensorInfo& tensorInfo)
38  {
39  armnn::armcomputetensorutils::BuildArmComputeTensor(m_Tensor, tensorInfo);
40  }
41 
42  ClTensorHandle(const TensorInfo& tensorInfo, DataLayout dataLayout)
43  {
44  armnn::armcomputetensorutils::BuildArmComputeTensor(m_Tensor, tensorInfo, dataLayout);
45  }
46 
47  arm_compute::CLTensor& GetTensor() override { return m_Tensor; }
48  arm_compute::CLTensor const& GetTensor() const override { return m_Tensor; }
49  virtual void Allocate() override {armnn::armcomputetensorutils::InitialiseArmComputeTensorEmpty(m_Tensor);}
50 
51  virtual void Manage() override
52  {
53  assert(m_MemoryGroup != nullptr);
54  m_MemoryGroup->manage(&m_Tensor);
55  }
56 
57  virtual const void* Map(bool blocking = true) const override
58  {
59  const_cast<arm_compute::CLTensor*>(&m_Tensor)->map(blocking);
60  return static_cast<const void*>(m_Tensor.buffer() + m_Tensor.info()->offset_first_element_in_bytes());
61  }
62 
63  virtual void Unmap() const override { const_cast<arm_compute::CLTensor*>(&m_Tensor)->unmap(); }
64 
65  virtual ITensorHandle* GetParent() const override { return nullptr; }
66 
67  virtual arm_compute::DataType GetDataType() const override
68  {
69  return m_Tensor.info()->data_type();
70  }
71 
72  virtual void SetMemoryGroup(const std::shared_ptr<arm_compute::IMemoryGroup>& memoryGroup) override
73  {
74  m_MemoryGroup = PolymorphicPointerDowncast<arm_compute::MemoryGroup>(memoryGroup);
75  }
76 
77  TensorShape GetStrides() const override
78  {
79  return armcomputetensorutils::GetStrides(m_Tensor.info()->strides_in_bytes());
80  }
81 
82  TensorShape GetShape() const override
83  {
84  return armcomputetensorutils::GetShape(m_Tensor.info()->tensor_shape());
85  }
86 
87 private:
88  // Only used for testing
89  void CopyOutTo(void* memory) const override
90  {
91  const_cast<armnn::ClTensorHandle*>(this)->Map(true);
92  switch(this->GetDataType())
93  {
94  case arm_compute::DataType::F32:
95  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
96  static_cast<float*>(memory));
97  break;
98  case arm_compute::DataType::U8:
99  case arm_compute::DataType::QASYMM8:
100  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
101  static_cast<uint8_t*>(memory));
102  break;
103  case arm_compute::DataType::QSYMM8_PER_CHANNEL:
104  case arm_compute::DataType::QASYMM8_SIGNED:
105  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
106  static_cast<int8_t*>(memory));
107  break;
108  case arm_compute::DataType::F16:
109  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
110  static_cast<armnn::Half*>(memory));
111  break;
112  case arm_compute::DataType::S16:
113  case arm_compute::DataType::QSYMM16:
114  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
115  static_cast<int16_t*>(memory));
116  break;
117  case arm_compute::DataType::S32:
118  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
119  static_cast<int32_t*>(memory));
120  break;
121  default:
122  {
123  throw armnn::UnimplementedException();
124  }
125  }
126  const_cast<armnn::ClTensorHandle*>(this)->Unmap();
127  }
128 
129  // Only used for testing
130  void CopyInFrom(const void* memory) override
131  {
132  this->Map(true);
133  switch(this->GetDataType())
134  {
135  case arm_compute::DataType::F32:
136  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const float*>(memory),
137  this->GetTensor());
138  break;
139  case arm_compute::DataType::U8:
140  case arm_compute::DataType::QASYMM8:
141  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const uint8_t*>(memory),
142  this->GetTensor());
143  break;
144  case arm_compute::DataType::F16:
145  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const armnn::Half*>(memory),
146  this->GetTensor());
147  break;
148  case arm_compute::DataType::S16:
149  case arm_compute::DataType::QSYMM8_PER_CHANNEL:
150  case arm_compute::DataType::QASYMM8_SIGNED:
151  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int8_t*>(memory),
152  this->GetTensor());
153  break;
154  case arm_compute::DataType::QSYMM16:
155  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int16_t*>(memory),
156  this->GetTensor());
157  break;
158  case arm_compute::DataType::S32:
159  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int32_t*>(memory),
160  this->GetTensor());
161  break;
162  default:
163  {
164  throw armnn::UnimplementedException();
165  }
166  }
167  this->Unmap();
168  }
169 
170  arm_compute::CLTensor m_Tensor;
171  std::shared_ptr<arm_compute::MemoryGroup> m_MemoryGroup;
172 };
173 
174 class ClSubTensorHandle : public IClTensorHandle
175 {
176 public:
177  ClSubTensorHandle(IClTensorHandle* parent,
178  const arm_compute::TensorShape& shape,
179  const arm_compute::Coordinates& coords)
180  : m_Tensor(&parent->GetTensor(), shape, coords)
181  {
182  parentHandle = parent;
183  }
184 
185  arm_compute::CLSubTensor& GetTensor() override { return m_Tensor; }
186  arm_compute::CLSubTensor const& GetTensor() const override { return m_Tensor; }
187 
188  virtual void Allocate() override {}
189  virtual void Manage() override {}
190 
191  virtual const void* Map(bool blocking = true) const override
192  {
193  const_cast<arm_compute::CLSubTensor*>(&m_Tensor)->map(blocking);
194  return static_cast<const void*>(m_Tensor.buffer() + m_Tensor.info()->offset_first_element_in_bytes());
195  }
196  virtual void Unmap() const override { const_cast<arm_compute::CLSubTensor*>(&m_Tensor)->unmap(); }
197 
198  virtual ITensorHandle* GetParent() const override { return parentHandle; }
199 
200  virtual arm_compute::DataType GetDataType() const override
201  {
202  return m_Tensor.info()->data_type();
203  }
204 
205  virtual void SetMemoryGroup(const std::shared_ptr<arm_compute::IMemoryGroup>&) override {}
206 
207  TensorShape GetStrides() const override
208  {
209  return armcomputetensorutils::GetStrides(m_Tensor.info()->strides_in_bytes());
210  }
211 
212  TensorShape GetShape() const override
213  {
214  return armcomputetensorutils::GetShape(m_Tensor.info()->tensor_shape());
215  }
216 
217 private:
218  // Only used for testing
219  void CopyOutTo(void* memory) const override
220  {
221  const_cast<ClSubTensorHandle*>(this)->Map(true);
222  switch(this->GetDataType())
223  {
224  case arm_compute::DataType::F32:
225  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
226  static_cast<float*>(memory));
227  break;
228  case arm_compute::DataType::U8:
229  case arm_compute::DataType::QASYMM8:
230  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
231  static_cast<uint8_t*>(memory));
232  break;
233  case arm_compute::DataType::F16:
234  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
235  static_cast<armnn::Half*>(memory));
236  break;
237  case arm_compute::DataType::QSYMM8_PER_CHANNEL:
238  case arm_compute::DataType::QASYMM8_SIGNED:
239  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
240  static_cast<int8_t*>(memory));
241  break;
242  case arm_compute::DataType::S16:
243  case arm_compute::DataType::QSYMM16:
244  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
245  static_cast<int16_t*>(memory));
246  break;
247  case arm_compute::DataType::S32:
248  armcomputetensorutils::CopyArmComputeITensorData(this->GetTensor(),
249  static_cast<int32_t*>(memory));
250  break;
251  default:
252  {
253  throw armnn::UnimplementedException();
254  }
255  }
256  const_cast<ClSubTensorHandle*>(this)->Unmap();
257  }
258 
259  // Only used for testing
260  void CopyInFrom(const void* memory) override
261  {
262  this->Map(true);
263  switch(this->GetDataType())
264  {
265  case arm_compute::DataType::F32:
266  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const float*>(memory),
267  this->GetTensor());
268  break;
269  case arm_compute::DataType::U8:
270  case arm_compute::DataType::QASYMM8:
271  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const uint8_t*>(memory),
272  this->GetTensor());
273  break;
274  case arm_compute::DataType::F16:
275  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const armnn::Half*>(memory),
276  this->GetTensor());
277  break;
278  case arm_compute::DataType::QSYMM8_PER_CHANNEL:
279  case arm_compute::DataType::QASYMM8_SIGNED:
280  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int8_t*>(memory),
281  this->GetTensor());
282  break;
283  case arm_compute::DataType::S16:
284  case arm_compute::DataType::QSYMM16:
285  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int16_t*>(memory),
286  this->GetTensor());
287  break;
288  case arm_compute::DataType::S32:
289  armcomputetensorutils::CopyArmComputeITensorData(static_cast<const int32_t*>(memory),
290  this->GetTensor());
291  break;
292  default:
293  {
294  throw armnn::UnimplementedException();
295  }
296  }
297  this->Unmap();
298  }
299 
300  mutable arm_compute::CLSubTensor m_Tensor;
301  ITensorHandle* parentHandle = nullptr;
302 };
303 
304 } // namespace armnn
armnn::ClTensorHandle::GetShape
TensorShape GetShape() const override
Get the number of elements for each dimension ordered from slowest iterating dimension to fastest ite...
Definition: ClTensorHandle.hpp:82
+
armnn::ClSubTensorHandle::GetStrides
TensorShape GetStrides() const override
Get the strides for each dimension ordered from largest to smallest where the smallest value is the s...
Definition: ClTensorHandle.hpp:207
+
armnn::DataLayout
DataLayout
Definition: Types.hpp:50
+
armnn::IClTensorHandle::SetMemoryGroup
virtual void SetMemoryGroup(const std::shared_ptr< arm_compute::IMemoryGroup > &memoryGroup)=0
+ + +
armnn::IClTensorHandle::GetDataType
virtual arm_compute::DataType GetDataType() const =0
+
armnn::ClSubTensorHandle::GetParent
virtual ITensorHandle * GetParent() const override
Get the parent tensor if this is a subtensor.
Definition: ClTensorHandle.hpp:198
+
armnn::Coordinates
std::array< unsigned int, MaxNumOfTensorDimensions > Coordinates
Definition: InternalTypes.hpp:14
+
armnn::ClTensorHandle::GetDataType
virtual arm_compute::DataType GetDataType() const override
Definition: ClTensorHandle.hpp:67
+ +
ArmComputeTensorUtils.hpp
+
armnn::ClTensorHandle::Map
virtual const void * Map(bool blocking=true) const override
Map the tensor data for access.
Definition: ClTensorHandle.hpp:57
+
armnn
Copyright (c) 2021 ARM Limited and Contributors.
Definition: 01_00_software_tools.dox:6
+ + +
armnn::ClTensorHandle::Unmap
virtual void Unmap() const override
Unmap the tensor data.
Definition: ClTensorHandle.hpp:63
+
armnn::ClTensorHandle::SetMemoryGroup
virtual void SetMemoryGroup(const std::shared_ptr< arm_compute::IMemoryGroup > &memoryGroup) override
Definition: ClTensorHandle.hpp:72
+
PolymorphicDowncast.hpp
+
armnn::ClSubTensorHandle::GetTensor
arm_compute::CLSubTensor const & GetTensor() const override
Definition: ClTensorHandle.hpp:186
+
armnn::ClTensorHandle::Allocate
virtual void Allocate() override
Indicate to the memory manager that this resource is no longer active.
Definition: ClTensorHandle.hpp:49
+
armnn::ClSubTensorHandle::Manage
virtual void Manage() override
Indicate to the memory manager that this resource is active.
Definition: ClTensorHandle.hpp:189
+
armnn::DataType
DataType
Definition: Types.hpp:32
+
armnn::ClTensorHandle::GetParent
virtual ITensorHandle * GetParent() const override
Get the parent tensor if this is a subtensor.
Definition: ClTensorHandle.hpp:65
+
armnn::ClSubTensorHandle::ClSubTensorHandle
ClSubTensorHandle(IClTensorHandle *parent, const arm_compute::TensorShape &shape, const arm_compute::Coordinates &coords)
Definition: ClTensorHandle.hpp:177
+ +
armnn::ClSubTensorHandle::GetDataType
virtual arm_compute::DataType GetDataType() const override
Definition: ClTensorHandle.hpp:200
+
armnn::ClTensorHandle::GetTensor
arm_compute::CLTensor const & GetTensor() const override
Definition: ClTensorHandle.hpp:48
+
armnn::ClTensorHandle::GetTensor
arm_compute::CLTensor & GetTensor() override
Definition: ClTensorHandle.hpp:47
+
armnn::ClTensorHandle::ClTensorHandle
ClTensorHandle(const TensorInfo &tensorInfo)
Definition: ClTensorHandle.hpp:37
+
armnn::ITensorHandle::CopyOutTo
virtual void CopyOutTo(void *memory) const =0
Testing support to be able to verify and set tensor data content.
+
armnn::ClTensorHandle::Manage
virtual void Manage() override
Indicate to the memory manager that this resource is active.
Definition: ClTensorHandle.hpp:51
+
Half.hpp
+
armnn::ITensorHandle::Map
virtual const void * Map(bool blocking=true) const =0
Map the tensor data for access.
+
armnn::ClSubTensorHandle::Map
virtual const void * Map(bool blocking=true) const override
Map the tensor data for access.
Definition: ClTensorHandle.hpp:191
+
armnn::ITensorHandle::Unmap
virtual void Unmap() const =0
Unmap the tensor data.
+ +
armnn::ClSubTensorHandle::Allocate
virtual void Allocate() override
Indicate to the memory manager that this resource is no longer active.
Definition: ClTensorHandle.hpp:188
+ +
armnn::ITensorHandle::CopyInFrom
virtual void CopyInFrom(const void *memory)=0
+
ArmComputeTensorHandle.hpp
+
armnn::ClTensorHandle::GetStrides
TensorShape GetStrides() const override
Get the strides for each dimension ordered from largest to smallest where the smallest value is the s...
Definition: ClTensorHandle.hpp:77
+
armnn::ClTensorHandle::ClTensorHandle
ClTensorHandle(const TensorInfo &tensorInfo, DataLayout dataLayout)
Definition: ClTensorHandle.hpp:42
+
armnn::ClSubTensorHandle::GetTensor
arm_compute::CLSubTensor & GetTensor() override
Definition: ClTensorHandle.hpp:185
+
armnn::IClTensorHandle::GetTensor
virtual arm_compute::ICLTensor & GetTensor()=0
+
armnn::ClSubTensorHandle::SetMemoryGroup
virtual void SetMemoryGroup(const std::shared_ptr< arm_compute::IMemoryGroup > &) override
Definition: ClTensorHandle.hpp:205
+
armnn::ClSubTensorHandle::Unmap
virtual void Unmap() const override
Unmap the tensor data.
Definition: ClTensorHandle.hpp:196
+
armnn::ClSubTensorHandle::GetShape
TensorShape GetShape() const override
Get the number of elements for each dimension ordered from slowest iterating dimension to fastest ite...
Definition: ClTensorHandle.hpp:212
+
+
+ + + + -- cgit v1.2.1