aboutsummaryrefslogtreecommitdiffstats
path: root/libavfilter/vf_scale_cuda.cu
blob: 3f3f40546d4bca8aad298398f9b99ffcf1c57ce7 (plain) (blame)
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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
/*
 * Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.
 *
 * 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.
 */

extern "C" {

__global__ void Subsample_Bilinear_uchar(cudaTextureObject_t uchar_tex,
                                    unsigned char *dst,
                                    int dst_width, int dst_height, int dst_pitch,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        int y0 = tex2D<unsigned char>(uchar_tex, xi-dx, yi-dy);
        int y1 = tex2D<unsigned char>(uchar_tex, xi+dx, yi-dy);
        int y2 = tex2D<unsigned char>(uchar_tex, xi-dx, yi+dy);
        int y3 = tex2D<unsigned char>(uchar_tex, xi+dx, yi+dy);
        dst[yo*dst_pitch+xo] = (unsigned char)((y0+y1+y2+y3+2) >> 2);
    }
}

__global__ void Subsample_Bilinear_uchar2(cudaTextureObject_t uchar2_tex,
                                    uchar2 *dst,
                                    int dst_width, int dst_height, int dst_pitch2,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        uchar2 c0 = tex2D<uchar2>(uchar2_tex, xi-dx, yi-dy);
        uchar2 c1 = tex2D<uchar2>(uchar2_tex, xi+dx, yi-dy);
        uchar2 c2 = tex2D<uchar2>(uchar2_tex, xi-dx, yi+dy);
        uchar2 c3 = tex2D<uchar2>(uchar2_tex, xi+dx, yi+dy);
        int2 uv;
        uv.x = ((int)c0.x+(int)c1.x+(int)c2.x+(int)c3.x+2) >> 2;
        uv.y = ((int)c0.y+(int)c1.y+(int)c2.y+(int)c3.y+2) >> 2;
        dst[yo*dst_pitch2+xo] = make_uchar2((unsigned char)uv.x, (unsigned char)uv.y);
    }
}

__global__ void Subsample_Bilinear_uchar4(cudaTextureObject_t uchar4_tex,
                                    uchar4 *dst,
                                    int dst_width, int dst_height, int dst_pitch,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        uchar4 c0 = tex2D<uchar4>(uchar4_tex, xi-dx, yi-dy);
        uchar4 c1 = tex2D<uchar4>(uchar4_tex, xi+dx, yi-dy);
        uchar4 c2 = tex2D<uchar4>(uchar4_tex, xi-dx, yi+dy);
        uchar4 c3 = tex2D<uchar4>(uchar4_tex, xi+dx, yi+dy);
        int4 res;
        res.x =  ((int)c0.x+(int)c1.x+(int)c2.x+(int)c3.x+2) >> 2;
        res.y =  ((int)c0.y+(int)c1.y+(int)c2.y+(int)c3.y+2) >> 2;
        res.z =  ((int)c0.z+(int)c1.z+(int)c2.z+(int)c3.z+2) >> 2;
        res.w =  ((int)c0.w+(int)c1.w+(int)c2.w+(int)c3.w+2) >> 2;
        dst[yo*dst_pitch+xo] = make_uchar4(
            (unsigned char)res.x, (unsigned char)res.y, (unsigned char)res.z, (unsigned char)res.w);
    }
}

__global__ void Subsample_Bilinear_ushort(cudaTextureObject_t ushort_tex,
                                    unsigned short *dst,
                                    int dst_width, int dst_height, int dst_pitch,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        int y0 = tex2D<unsigned short>(ushort_tex, xi-dx, yi-dy);
        int y1 = tex2D<unsigned short>(ushort_tex, xi+dx, yi-dy);
        int y2 = tex2D<unsigned short>(ushort_tex, xi-dx, yi+dy);
        int y3 = tex2D<unsigned short>(ushort_tex, xi+dx, yi+dy);
        dst[yo*dst_pitch+xo] = (unsigned short)((y0+y1+y2+y3+2) >> 2);
    }
}

__global__ void Subsample_Bilinear_ushort2(cudaTextureObject_t ushort2_tex,
                                    ushort2 *dst,
                                    int dst_width, int dst_height, int dst_pitch2,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        ushort2 c0 = tex2D<ushort2>(ushort2_tex, xi-dx, yi-dy);
        ushort2 c1 = tex2D<ushort2>(ushort2_tex, xi+dx, yi-dy);
        ushort2 c2 = tex2D<ushort2>(ushort2_tex, xi-dx, yi+dy);
        ushort2 c3 = tex2D<ushort2>(ushort2_tex, xi+dx, yi+dy);
        int2 uv;
        uv.x = ((int)c0.x+(int)c1.x+(int)c2.x+(int)c3.x+2) >> 2;
        uv.y = ((int)c0.y+(int)c1.y+(int)c2.y+(int)c3.y+2) >> 2;
        dst[yo*dst_pitch2+xo] = make_ushort2((unsigned short)uv.x, (unsigned short)uv.y);
    }
}

__global__ void Subsample_Bilinear_ushort4(cudaTextureObject_t ushort4_tex,
                                    ushort4 *dst,
                                    int dst_width, int dst_height, int dst_pitch,
                                    int src_width, int src_height)
{
    int xo = blockIdx.x * blockDim.x + threadIdx.x;
    int yo = blockIdx.y * blockDim.y + threadIdx.y;

    if (yo < dst_height && xo < dst_width)
    {
        float hscale = (float)src_width / (float)dst_width;
        float vscale = (float)src_height / (float)dst_height;
        float xi = (xo + 0.5f) * hscale;
        float yi = (yo + 0.5f) * vscale;
        // 3-tap filter weights are {wh,1.0,wh} and {wv,1.0,wv}
        float wh = min(max(0.5f * (hscale - 1.0f), 0.0f), 1.0f);
        float wv = min(max(0.5f * (vscale - 1.0f), 0.0f), 1.0f);
        // Convert weights to two bilinear weights -> {wh,1.0,wh} -> {wh,0.5,0} + {0,0.5,wh}
        float dx = wh / (0.5f + wh);
        float dy = wv / (0.5f + wv);
        ushort4 c0 = tex2D<ushort4>(ushort4_tex, xi-dx, yi-dy);
        ushort4 c1 = tex2D<ushort4>(ushort4_tex, xi+dx, yi-dy);
        ushort4 c2 = tex2D<ushort4>(ushort4_tex, xi-dx, yi+dy);
        ushort4 c3 = tex2D<ushort4>(ushort4_tex, xi+dx, yi+dy);
        int4 res;
        res.x =  ((int)c0.x+(int)c1.x+(int)c2.x+(int)c3.x+2) >> 2;
        res.y =  ((int)c0.y+(int)c1.y+(int)c2.y+(int)c3.y+2) >> 2;
        res.z =  ((int)c0.z+(int)c1.z+(int)c2.z+(int)c3.z+2) >> 2;
        res.w =  ((int)c0.w+(int)c1.w+(int)c2.w+(int)c3.w+2) >> 2;
        dst[yo*dst_pitch+xo] = make_ushort4(
            (unsigned short)res.x, (unsigned short)res.y, (unsigned short)res.z, (unsigned short)res.w);
    }
}

}