Hugging Face's logo

camenduru
/
ncnn

Model card Files
xet
Community
ncnn
File size: 4,212 Bytes
be903e2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
 
// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2020 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#include "clip_riscv.h"

#if __riscv_vector
#include <riscv_vector.h>
#include "rvv_mathfun.h"
#include "rvv_mathfun_fp16s.h"
#endif // __riscv_vector

namespace ncnn {

Clip_riscv::Clip_riscv()
{
#if __riscv_vector
    support_packing = true;
#if __riscv_zfh
    support_fp16_storage = true;
#endif
#endif // __riscv_vector
}

int Clip_riscv::forward_inplace(Mat& bottom_top_blob, const Option& opt) const
{
#if __riscv_vector && __riscv_zfh
    int elembits = bottom_top_blob.elembits();

    if (opt.use_fp16_storage && elembits == 16)
    {
        if (opt.use_fp16_arithmetic)
            return forward_inplace_fp16sa(bottom_top_blob, opt);
        else
            return forward_inplace_fp16s(bottom_top_blob, opt);
    }
#endif

    int w = bottom_top_blob.w;
    int h = bottom_top_blob.h;
    int d = bottom_top_blob.d;
    int channels = bottom_top_blob.c;
    int elempack = bottom_top_blob.elempack;
    int size = w * h * d * elempack;

    #pragma omp parallel for num_threads(opt.num_threads)
    for (int q = 0; q < channels; q++)
    {
        float* ptr = bottom_top_blob.channel(q);

#if __riscv_vector
        int n = size;
        while (n > 0)
        {
            size_t vl = vsetvl_e32m8(n);

            vfloat32m8_t _p = vle32_v_f32m8(ptr, vl);
            _p = vfmax_vf_f32m8(_p, min, vl);
            _p = vfmin_vf_f32m8(_p, max, vl);
            vse32_v_f32m8(ptr, _p, vl);

            ptr += vl;
            n -= vl;
        }
#else  // __riscv_vector
        for (int i = 0; i < size; i++)
        {
            if (*ptr < min)
                *ptr = min;

            if (*ptr > max)
                *ptr = max;

            ptr++;
        }
#endif // __riscv_vector
    }

    return 0;
}

#if __riscv_vector && __riscv_zfh
int Clip_riscv::forward_inplace_fp16s(Mat& bottom_top_blob, const Option& opt) const
{
    int w = bottom_top_blob.w;
    int h = bottom_top_blob.h;
    int d = bottom_top_blob.d;
    int channels = bottom_top_blob.c;
    int elempack = bottom_top_blob.elempack;
    int size = w * h * d * elempack;

    #pragma omp parallel for num_threads(opt.num_threads)
    for (int q = 0; q < channels; q++)
    {
        __fp16* ptr = bottom_top_blob.channel(q);

        int n = size;
        while (n > 0)
        {
            size_t vl = vsetvl_e16m4(n);

            vfloat32m8_t _p = vfwcvt_f_f_v_f32m8(vle16_v_f16m4(ptr, vl), vl);
            _p = vfmax_vf_f32m8(_p, min, vl);
            _p = vfmin_vf_f32m8(_p, max, vl);
            vse16_v_f16m4(ptr, vfncvt_f_f_w_f16m4(_p, vl), vl);

            ptr += vl;
            n -= vl;
        }
    }

    return 0;
}

int Clip_riscv::forward_inplace_fp16sa(Mat& bottom_top_blob, const Option& opt) const
{
    int w = bottom_top_blob.w;
    int h = bottom_top_blob.h;
    int d = bottom_top_blob.d;
    int channels = bottom_top_blob.c;
    int elempack = bottom_top_blob.elempack;
    int size = w * h * d * elempack;

    #pragma omp parallel for num_threads(opt.num_threads)
    for (int q = 0; q < channels; q++)
    {
        __fp16* ptr = bottom_top_blob.channel(q);

        int n = size;
        while (n > 0)
        {
            size_t vl = vsetvl_e16m8(n);

            vfloat16m8_t _p = vle16_v_f16m8(ptr, vl);
            _p = vfmax_vf_f16m8(_p, min, vl);
            _p = vfmin_vf_f16m8(_p, max, vl);
            vse16_v_f16m8(ptr, _p, vl);

            ptr += vl;
            n -= vl;
        }
    }

    return 0;
}
#endif // __riscv_vector && __riscv_zfh

} //namespace ncnn