2 * Copyright (c) 2016, 2017 ARM Limited.
4 * SPDX-License-Identifier: MIT
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:
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
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
24 #include "arm_compute/core/NEON/kernels/NEErodeKernel.h"
26 #include "arm_compute/core/Helpers.h"
27 #include "arm_compute/core/ITensor.h"
28 #include "arm_compute/core/NEON/INEKernel.h"
29 #include "arm_compute/core/TensorInfo.h"
30 #include "arm_compute/core/Validate.h"
36 using namespace arm_compute;
41 } // namespace arm_compute
43 BorderSize NEErodeKernel::border_size() const
48 void NEErodeKernel::configure(const ITensor *input, ITensor *output, bool border_undefined)
53 constexpr unsigned int num_elems_processed_per_iteration = 8;
54 constexpr unsigned int num_elems_read_per_iteration = 16;
55 constexpr unsigned int num_elems_written_per_iteration = 8;
56 constexpr unsigned int num_rows_read_per_iteration = 3;
58 // Configure kernel window
59 Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration), border_undefined, border_size());
60 AccessWindowHorizontal output_access(output->info(), 0, num_elems_written_per_iteration);
61 AccessWindowRectangle input_access(input->info(), -border_size().left, -border_size().top, num_elems_read_per_iteration, num_rows_read_per_iteration);
63 update_window_and_padding(win, input_access, output_access);
65 output_access.set_valid_region(win, input->info()->valid_region(), border_undefined, border_size());
67 INEKernel::configure(win);
70 void NEErodeKernel::run(const Window &window)
72 ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
73 ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INESimpleKernel::window(), window);
75 Iterator in(_input, window);
76 Iterator out(_output, window);
78 const size_t in_stride = _input->info()->strides_in_bytes()[1];
80 execute_window_loop(window, [&](const Coordinates &)
82 uint8_t *in_ptr = in.ptr() - 1;
83 const uint8x16_t top_data = vld1q_u8(in_ptr - in_stride);
84 const uint8x16_t mid_data = vld1q_u8(in_ptr);
85 const uint8x16_t bot_data = vld1q_u8(in_ptr + in_stride);
87 uint8x8_t top_high_data = vget_high_u8(top_data);
88 uint8x8_t top_low_data = vget_low_u8(top_data);
90 uint8x8_t mid_high_data = vget_high_u8(mid_data);
91 uint8x8_t mid_low_data = vget_low_u8(mid_data);
93 uint8x8_t bot_high_data = vget_high_u8(bot_data);
94 uint8x8_t bot_low_data = vget_low_u8(bot_data);
99 p1 = vext_u8(top_low_data, top_high_data, 1);
100 p0 = vmin_u8(p0, p1);
102 p1 = vext_u8(top_low_data, top_high_data, 2);
103 p0 = vmin_u8(p0, p1);
106 p0 = vmin_u8(p0, p1);
108 p1 = vext_u8(mid_low_data, mid_high_data, 1);
109 p0 = vmin_u8(p0, p1);
111 p1 = vext_u8(mid_low_data, mid_high_data, 2);
112 p0 = vmin_u8(p0, p1);
115 p0 = vmin_u8(p0, p1);
117 p1 = vext_u8(bot_low_data, bot_high_data, 1);
118 p0 = vmin_u8(p0, p1);
120 p1 = vext_u8(bot_low_data, bot_high_data, 2);
121 p0 = vmin_u8(p0, p1);
123 vst1_u8(out.ptr(), p0);