1 /*
2 * Copyright (c) 2018-2021 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/gpu/cl/kernels/ClFillKernel.h"
25
26 #include "arm_compute/core/CL/CLHelpers.h"
27 #include "arm_compute/core/CL/CLKernelLibrary.h"
28 #include "arm_compute/core/CL/ICLTensor.h"
29 #include "arm_compute/core/Helpers.h"
30 #include "arm_compute/core/TensorInfo.h"
31 #include "arm_compute/core/Utils.h"
32 #include "arm_compute/core/Validate.h"
33 #include "src/core/CL/CLValidate.h"
34 #include "src/core/helpers/AutoConfiguration.h"
35 #include "src/core/helpers/WindowHelpers.h"
36 #include "support/Cast.h"
37 #include "support/StringSupport.h"
38
39 namespace arm_compute
40 {
41 namespace opencl
42 {
43 namespace kernels
44 {
ClFillKernel()45 ClFillKernel::ClFillKernel()
46 {
47 _type = CLKernelType::ELEMENTWISE;
48 }
49
configure(const CLCompileContext & compile_context,ITensorInfo * tensor,const PixelValue & constant_value,Window * window)50 void ClFillKernel::configure(const CLCompileContext &compile_context, ITensorInfo *tensor,
51 const PixelValue &constant_value,
52 Window *window)
53 {
54 ARM_COMPUTE_ERROR_ON_NULLPTR(tensor);
55 ARM_COMPUTE_ERROR_THROW_ON(validate(tensor, constant_value, window));
56
57 const DataType data_type = tensor->data_type();
58 const int vec_size_x = 16 / tensor->element_size();
59
60 // Create and update the window (if needed)
61 _full_window = calculate_max_window(*tensor);
62 Window win = _full_window;
63 if(window != nullptr)
64 {
65 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(win, *window);
66 win = *window;
67 }
68
69 const int output_width_x = win.num_iterations(0);
70 const bool multi_access_x = output_width_x >= vec_size_x;
71 const bool remainder_x = output_width_x % vec_size_x > 0;
72
73 if(multi_access_x)
74 {
75 win.set(Window::DimX, Window::Dimension(win.x().start(), ceil_to_multiple(win.x().end(), vec_size_x), vec_size_x));
76 }
77 ICLKernel::configure_internal(win);
78
79 // Create kernel
80 CLBuildOptions build_opts;
81 build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
82 build_opts.add_option("-DCONSTANT_VALUE=" + string_from_pixel_value(constant_value, data_type));
83 build_opts.add_option_if(multi_access_x, "-DVEC_SIZE=" + support::cpp11::to_string(vec_size_x));
84 build_opts.add_option_if(multi_access_x && remainder_x, "-DLAST_ACCESSED_X=" + support::cpp11::to_string(std::max<int>(output_width_x - vec_size_x, 0)));
85 _kernel = create_kernel(compile_context, "memset", build_opts.options());
86 }
87
validate(const ITensorInfo * tensor,const PixelValue & constant_value,Window * window)88 Status ClFillKernel::validate(const ITensorInfo *tensor, const PixelValue &constant_value, Window *window)
89 {
90 ARM_COMPUTE_UNUSED(tensor);
91 ARM_COMPUTE_UNUSED(constant_value);
92 if(window != nullptr)
93 {
94 ARM_COMPUTE_RETURN_ERROR_ON(window->x().step() != 1);
95 }
96 return Status{};
97 }
98
run_op(ITensorPack & tensors,const Window & window,cl::CommandQueue & queue)99 void ClFillKernel::run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
100 {
101 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
102 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
103
104 const auto tensor = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
105
106 // Collapse all the batches on the third
107 Window collapsed = window.collapse_if_possible(_full_window, Window::DimZ);
108 Window slice = collapsed.first_slice_window_3D();
109
110 do
111 {
112 unsigned int idx = 0;
113 add_3D_tensor_argument(idx, tensor, slice);
114 enqueue(queue, *this, slice, lws_hint());
115 }
116 while(collapsed.slide_window_slice_3D(slice));
117 }
118 } // namespace kernels
119 } // namespace opencl
120 } // namespace arm_compute
121