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 */
24 #include "src/core/CL/kernels/CLScharr3x3Kernel.h"
25
26 #include "arm_compute/core/CL/CLKernelLibrary.h"
27 #include "arm_compute/core/CL/ICLTensor.h"
28 #include "arm_compute/core/CL/OpenCL.h"
29 #include "arm_compute/core/Helpers.h"
30 #include "arm_compute/core/Validate.h"
31 #include "src/core/helpers/WindowHelpers.h"
32
33 #include <set>
34 #include <string>
35
36 using namespace arm_compute;
37
CLScharr3x3Kernel()38 CLScharr3x3Kernel::CLScharr3x3Kernel()
39 : _run_scharr_x(false), _run_scharr_y(false), _input(nullptr), _output_x(nullptr), _output_y(nullptr)
40 {
41 }
42
border_size() const43 BorderSize CLScharr3x3Kernel::border_size() const
44 {
45 return BorderSize(1);
46 }
47
configure(const ICLTensor * input,ICLTensor * output_x,ICLTensor * output_y,bool border_undefined)48 void CLScharr3x3Kernel::configure(const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
49 {
50 configure(CLKernelLibrary::get().get_compile_context(), input, output_x, output_y, border_undefined);
51 }
52
configure(const CLCompileContext & compile_context,const ICLTensor * input,ICLTensor * output_x,ICLTensor * output_y,bool border_undefined)53 void CLScharr3x3Kernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined)
54 {
55 ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8);
56 ARM_COMPUTE_ERROR_ON((output_x == nullptr) && (output_y == nullptr));
57
58 _run_scharr_x = output_x != nullptr;
59 _run_scharr_y = output_y != nullptr;
60
61 if(_run_scharr_x)
62 {
63 ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_x, 1, DataType::S16);
64 }
65
66 if(_run_scharr_y)
67 {
68 ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output_y, 1, DataType::S16);
69 }
70
71 _input = input;
72 _output_x = output_x;
73 _output_y = output_y;
74
75 // Set build options
76 std::set<std::string> build_opts;
77
78 if(_run_scharr_x)
79 {
80 build_opts.insert("-DGRAD_X");
81 }
82
83 if(_run_scharr_y)
84 {
85 build_opts.insert("-DGRAD_Y");
86 }
87
88 // Create kernel
89 _kernel = create_kernel(compile_context, "scharr3x3", build_opts);
90
91 // Configure kernel window
92 constexpr unsigned int num_elems_processed_per_iteration = 8;
93 constexpr unsigned int num_elems_read_per_iteration = 16;
94 constexpr unsigned int num_elems_written_per_iteration = 8;
95 constexpr unsigned int num_rows_read_per_iteration = 3;
96
97 Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
98
99 AccessWindowRectangle input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
100 AccessWindowHorizontal output_x_access(output_x == nullptr ? nullptr : output_x->info(), 0, num_elems_written_per_iteration);
101 AccessWindowHorizontal output_y_access(output_y == nullptr ? nullptr : output_y->info(), 0, num_elems_written_per_iteration);
102
103 update_window_and_padding(win, input_access, output_x_access, output_y_access);
104
105 output_x_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
106 output_y_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
107
108 ICLKernel::configure_internal(win);
109 }
110
run(const Window & window,cl::CommandQueue & queue)111 void CLScharr3x3Kernel::run(const Window &window, cl::CommandQueue &queue)
112 {
113 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
114 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
115
116 Window slice = window.first_slice_window_2D();
117 do
118 {
119 unsigned int idx = 0;
120 add_2D_tensor_argument(idx, _input, slice);
121 add_2D_tensor_argument_if((_run_scharr_x), idx, _output_x, slice);
122 add_2D_tensor_argument_if((_run_scharr_y), idx, _output_y, slice);
123
124 enqueue(queue, *this, slice, lws_hint());
125 }
126 while(window.slide_window_slice_2D(slice));
127 }
128