aboutsummaryrefslogtreecommitdiff
path: root/src/cpu/kernels/scale/sve/qasymm8.cpp
blob: d45a69e43b13b69d09f15ec91e242024437aa623 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
/*
 * Copyright (c) 2021-2022 Arm Limited.
 *
 * SPDX-License-Identifier: MIT
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to
 * deal in the Software without restriction, including without limitation the
 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
 * sell copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include "arm_compute/core/Helpers.h"
#include "arm_compute/core/ITensorPack.h"
#include "arm_compute/core/Window.h"
#include "src/core/NEON/NEMath.h"
#include "src/core/NEON/wrapper/wrapper.h"
#include "src/core/helpers/ScaleHelpers.h"
#include "src/core/helpers/ScaleHelpers.h"
#include "src/core/utils/ScaleUtils.h"
#include "support/Rounding.h"

#include <arm_sve.h>
#include <cmath>
#include <cstddef>

namespace arm_compute
{
namespace
{
void qasymm8_sve_scale_nearest(const ITensor *src, ITensor *dst, const ITensor *offsets,
                               float sampling_offset, bool align_corners, const Window &window)
{
    const size_t in_stride_c  = src->info()->dimension(0) + src->info()->padding().left + src->info()->padding().right;
    const size_t in_stride_w  = src->info()->dimension(1) + src->info()->padding().top + src->info()->padding().bottom;
    const size_t in_stride_wc = in_stride_w * in_stride_c;
    const size_t in_dim_h     = src->info()->dimension(2);

    // Compute the ratio between source height and destination height
    const auto hr             = scale_utils::calculate_resize_ratio(in_dim_h, dst->info()->dimension(2), align_corners);
    const auto window_start_x = static_cast<int32_t>(window.x().start());
    const auto window_end_x   = static_cast<int32_t>(window.x().end());

    Window win(window);
    win.set(Window::DimX, Window::Dimension(0, 1, 1));
    Iterator out(dst, win);

    const uint8_t     *in_ptr_start        = src->buffer() + src->info()->offset_first_element_in_bytes();
    const unsigned int in_stride_bytes_hwc = src->info()->strides_in_bytes()[3];

    execute_window_loop(win, [&](const Coordinates & id)
    {
        const int32_t offset     = *reinterpret_cast<const int32_t *>(offsets->ptr_to_element(Coordinates(id.y(), id.z()))) * in_stride_c;
        const auto    in_hi      = static_cast<int>(align_corners ? utils::rounding::round_half_away_from_zero((id.z() + sampling_offset) * hr) : std::floor((id.z() + sampling_offset) * hr));
        const int     offset_row = in_hi * in_stride_wc;
        const auto    in_ptr     = reinterpret_cast<const uint8_t *>(in_ptr_start + in_stride_bytes_hwc * id[3]);
        const auto    out_ptr    = reinterpret_cast<uint8_t *>(out.ptr());

        // Compute S elements per iteration
        int      x  = window_start_x;
        svbool_t pg = svwhilelt_b8(x, window_end_x);
        do
        {
            // Store results
            svst1_u8(pg, out_ptr + x, svld1_u8(pg, in_ptr + offset + offset_row + x));

            x += svcntw();
            pg = svwhilelt_b8(x, window_end_x);
        }
        while(svptest_any(svptrue_b8(), pg));
    },
    out);
}
}
namespace cpu
{
void qasymm8_sve_scale(const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy,
                       InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset,
                       bool align_corners, const Window &window)
{
    ARM_COMPUTE_UNUSED(dx, dy, border_mode, constant_border_value);
    if(policy == InterpolationPolicy::NEAREST_NEIGHBOR)
    {
        qasymm8_sve_scale_nearest(src, dst, offsets, sampling_offset, align_corners, window);
    }
    else
    {
        ARM_COMPUTE_ERROR("Not Implemented");
    }
}
} // namespace cpu
} // namespace arm_compute