xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/kernels/CpuFillKernel.cpp (revision c217d954acce2dbc11938adb493fc0abd69584f3)
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/cpu/kernels/CpuFillKernel.h"
25 
26 #include "arm_compute/core/Helpers.h"
27 #include "arm_compute/core/ITensor.h"
28 #include "arm_compute/core/Utils.h"
29 #include "arm_compute/core/Validate.h"
30 #include "arm_compute/core/Window.h"
31 #include "src/core/helpers/AutoConfiguration.h"
32 #include "src/core/helpers/WindowHelpers.h"
33 
34 namespace arm_compute
35 {
36 namespace cpu
37 {
38 namespace kernels
39 {
configure(const ITensorInfo * tensor,const PixelValue & constant_value)40 void CpuFillKernel::configure(const ITensorInfo *tensor, const PixelValue &constant_value)
41 {
42     ARM_COMPUTE_ERROR_ON_NULLPTR(tensor);
43     _constant_value = constant_value;
44 
45     // Configure kernel window
46     Window win = calculate_max_window(*tensor, Steps());
47     ICpuKernel::configure(win);
48 }
49 
run_op(ITensorPack & tensors,const Window & window,const ThreadInfo & info)50 void CpuFillKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info)
51 {
52     ARM_COMPUTE_UNUSED(info);
53     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
54     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICpuKernel::window(), window);
55 
56     auto inout = tensors.get_tensor(TensorType::ACL_SRC_DST);
57 
58     // Collapse all the batches on the third dimension
59     bool   has_collapsed = true;
60     Window collapsed     = window.collapse_if_possible(window, Window::DimZ, &has_collapsed);
61     ARM_COMPUTE_ERROR_ON(!has_collapsed);
62 
63     uint8_t *const start_valid_region = inout->ptr_to_element(inout->info()->valid_region().anchor);
64     const auto     window_width       = static_cast<int>(collapsed.x().end()) - static_cast<int>(collapsed.x().start());
65     const size_t   element_size       = inout->info()->element_size();
66 
67     // Unroll X dimension
68     collapsed.set(Window::DimX, Window::Dimension(0, 1, 1));
69 
70     Iterator tensor_it(inout, collapsed);
71     execute_window_loop(collapsed, [&](const Coordinates &)
72     {
73         uint8_t *base_addr = start_valid_region + tensor_it.offset();
74         // Set memory
75         for(int i = 0; i < window_width; ++i)
76         {
77             std::memcpy(base_addr + i * element_size, &_constant_value.value, element_size);
78         }
79 
80     },
81     tensor_it);
82 }
83 
name() const84 const char *CpuFillKernel::name() const
85 {
86     return "CpuFillKernel";
87 }
88 } // namespace kernels
89 } // namespace cpu
90 } // namespace arm_compute
91