Compute Library
 21.02
CLGaussian5x5.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016-2020 Arm Limited.
3  *
4  * SPDX-License-Identifier: MIT
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to
8  * deal in the Software without restriction, including without limitation the
9  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10  * sell copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in all
14  * copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
25 
34 
35 #include <utility>
36 
37 using namespace arm_compute;
38 
39 CLGaussian5x5::CLGaussian5x5(std::shared_ptr<IMemoryManager> memory_manager)
40  : _memory_group(std::move(memory_manager)),
41  _kernel_hor(std::make_unique<CLGaussian5x5HorKernel>()),
42  _kernel_vert(std::make_unique<CLGaussian5x5VertKernel>()),
43  _border_handler(std::make_unique<CLFillBorderKernel>()),
44  _tmp()
45 {
46 }
47 
49 
50 void CLGaussian5x5::configure(ICLTensor *input, ICLTensor *output, BorderMode border_mode, uint8_t constant_border_value)
51 {
52  configure(CLKernelLibrary::get().get_compile_context(), input, output, border_mode, constant_border_value);
53 }
54 
55 void CLGaussian5x5::configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output, BorderMode border_mode, uint8_t constant_border_value)
56 {
58 
59  _tmp.allocator()->init(TensorInfo(input->info()->tensor_shape(), 1, DataType::U16));
60 
61  // Manage intermediate buffers
62  _memory_group.manage(&_tmp);
63 
64  // Configure kernels
65  _kernel_hor->configure(compile_context, input, &_tmp, border_mode == BorderMode::UNDEFINED);
66  _kernel_vert->configure(compile_context, &_tmp, output, border_mode == BorderMode::UNDEFINED);
67  _border_handler->configure(compile_context, input, _kernel_hor->border_size(), border_mode, PixelValue(constant_border_value));
68 
69  // Allocate intermediate buffers
70  _tmp.allocator()->allocate();
71 }
72 
74 {
75  CLScheduler::get().enqueue(*_border_handler, false);
76 
77  MemoryGroupResourceScope scope_mg(_memory_group);
78 
79  CLScheduler::get().enqueue(*_kernel_hor, false);
80  CLScheduler::get().enqueue(*_kernel_vert);
81 }
BorderMode
Methods available to handle borders.
Definition: Types.h:265
void configure(ICLTensor *input, ICLTensor *output, BorderMode border_mode, uint8_t constant_border_value=0)
Initialise the function&#39;s source, destinations and border mode.
Class describing the value of a pixel for any image format.
Definition: PixelValue.h:34
static CLScheduler & get()
Access the scheduler singleton.
1 channel, 1 U8 per channel
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
CLTensorAllocator * allocator()
Return a pointer to the tensor&#39;s allocator.
Definition: CLTensor.cpp:61
1 channel, 1 U16 per channel
void run() override
Run the kernels contained in the function.
void init(const TensorInfo &input, size_t alignment=0)
Initialize a tensor based on the passed TensorInfo.
Copyright (c) 2017-2021 Arm Limited.
void manage(IMemoryManageable *obj) override
Sets a object to be managed by the given memory group.
Definition: MemoryGroup.h:79
Interface to enqueue OpenCL kernels and get/set the OpenCL CommandQueue and ICLTuner.
Interface for the kernel to run the vertical pass of 5x5 Gaussian filter on a tensor.
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
CLGaussian5x5(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Default Constructor.
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
void enqueue(ICLKernel &kernel, bool flush=true)
Schedule the execution of the passed kernel if possible.
#define ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:790
CLCompileContext class.
void allocate() override
Allocate size specified by TensorInfo of OpenCL memory.
Interface for filling the border of a kernel.
Memory group resources scope handling class.
Definition: IMemoryGroup.h:82
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
Borders are left undefined.
Store the tensor&#39;s metadata.
Definition: TensorInfo.h:45
~CLGaussian5x5()
Default destructor.
Interface for the kernel to run the horizontal pass of 5x5 Gaussian filter on a tensor.