File size: 8,120 Bytes
8b7c501
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
// Auto-generated file. Do not edit!
//   Template: src/f32-dwconv/unipass-scalar.c.in
//   Generator: tools/xngen
//
// Copyright 2019 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.

#include <assert.h>

#include <xnnpack/dwconv.h>
#include <xnnpack/math.h>


void xnn_f32_dwconv_ukernel_25p1c__scalar(
    size_t channels,
    size_t output_width,
    const float** input,
    const float* weights,
    float* output,
    intptr_t input_stride,
    size_t output_increment,
    size_t input_offset,
    const float* zero,
    const union xnn_f32_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{
  assert(channels != 0);
  assert(output_width != 0);

  do {
    const float* i0 = input[0];
    assert(i0 != NULL);
    if XNN_UNPREDICTABLE(i0 != zero) {
      i0 = (const float*) ((uintptr_t) i0 + input_offset);
    }
    const float* i1 = input[1];
    assert(i1 != NULL);
    if XNN_UNPREDICTABLE(i1 != zero) {
      i1 = (const float*) ((uintptr_t) i1 + input_offset);
    }
    const float* i2 = input[2];
    assert(i2 != NULL);
    if XNN_UNPREDICTABLE(i2 != zero) {
      i2 = (const float*) ((uintptr_t) i2 + input_offset);
    }
    const float* i3 = input[3];
    assert(i3 != NULL);
    if XNN_UNPREDICTABLE(i3 != zero) {
      i3 = (const float*) ((uintptr_t) i3 + input_offset);
    }
    const float* i4 = input[4];
    assert(i4 != NULL);
    if XNN_UNPREDICTABLE(i4 != zero) {
      i4 = (const float*) ((uintptr_t) i4 + input_offset);
    }
    const float* i5 = input[5];
    assert(i5 != NULL);
    if XNN_UNPREDICTABLE(i5 != zero) {
      i5 = (const float*) ((uintptr_t) i5 + input_offset);
    }
    const float* i6 = input[6];
    assert(i6 != NULL);
    if XNN_UNPREDICTABLE(i6 != zero) {
      i6 = (const float*) ((uintptr_t) i6 + input_offset);
    }
    const float* i7 = input[7];
    assert(i7 != NULL);
    if XNN_UNPREDICTABLE(i7 != zero) {
      i7 = (const float*) ((uintptr_t) i7 + input_offset);
    }
    const float* i8 = input[8];
    assert(i8 != NULL);
    if XNN_UNPREDICTABLE(i8 != zero) {
      i8 = (const float*) ((uintptr_t) i8 + input_offset);
    }
    const float* i9 = input[9];
    assert(i9 != NULL);
    if XNN_UNPREDICTABLE(i9 != zero) {
      i9 = (const float*) ((uintptr_t) i9 + input_offset);
    }
    const float* i10 = input[10];
    assert(i10 != NULL);
    if XNN_UNPREDICTABLE(i10 != zero) {
      i10 = (const float*) ((uintptr_t) i10 + input_offset);
    }
    const float* i11 = input[11];
    assert(i11 != NULL);
    if XNN_UNPREDICTABLE(i11 != zero) {
      i11 = (const float*) ((uintptr_t) i11 + input_offset);
    }
    const float* i12 = input[12];
    assert(i12 != NULL);
    if XNN_UNPREDICTABLE(i12 != zero) {
      i12 = (const float*) ((uintptr_t) i12 + input_offset);
    }
    const float* i13 = input[13];
    assert(i13 != NULL);
    if XNN_UNPREDICTABLE(i13 != zero) {
      i13 = (const float*) ((uintptr_t) i13 + input_offset);
    }
    const float* i14 = input[14];
    assert(i14 != NULL);
    if XNN_UNPREDICTABLE(i14 != zero) {
      i14 = (const float*) ((uintptr_t) i14 + input_offset);
    }
    const float* i15 = input[15];
    assert(i15 != NULL);
    if XNN_UNPREDICTABLE(i15 != zero) {
      i15 = (const float*) ((uintptr_t) i15 + input_offset);
    }
    const float* i16 = input[16];
    assert(i16 != NULL);
    if XNN_UNPREDICTABLE(i16 != zero) {
      i16 = (const float*) ((uintptr_t) i16 + input_offset);
    }
    const float* i17 = input[17];
    assert(i17 != NULL);
    if XNN_UNPREDICTABLE(i17 != zero) {
      i17 = (const float*) ((uintptr_t) i17 + input_offset);
    }
    const float* i18 = input[18];
    assert(i18 != NULL);
    if XNN_UNPREDICTABLE(i18 != zero) {
      i18 = (const float*) ((uintptr_t) i18 + input_offset);
    }
    const float* i19 = input[19];
    assert(i19 != NULL);
    if XNN_UNPREDICTABLE(i19 != zero) {
      i19 = (const float*) ((uintptr_t) i19 + input_offset);
    }
    const float* i20 = input[20];
    assert(i20 != NULL);
    if XNN_UNPREDICTABLE(i20 != zero) {
      i20 = (const float*) ((uintptr_t) i20 + input_offset);
    }
    const float* i21 = input[21];
    assert(i21 != NULL);
    if XNN_UNPREDICTABLE(i21 != zero) {
      i21 = (const float*) ((uintptr_t) i21 + input_offset);
    }
    const float* i22 = input[22];
    assert(i22 != NULL);
    if XNN_UNPREDICTABLE(i22 != zero) {
      i22 = (const float*) ((uintptr_t) i22 + input_offset);
    }
    const float* i23 = input[23];
    assert(i23 != NULL);
    if XNN_UNPREDICTABLE(i23 != zero) {
      i23 = (const float*) ((uintptr_t) i23 + input_offset);
    }
    const float* i24 = input[24];
    assert(i24 != NULL);
    if XNN_UNPREDICTABLE(i24 != zero) {
      i24 = (const float*) ((uintptr_t) i24 + input_offset);
    }
    input = (const float**) ((uintptr_t) input + input_stride);

    size_t c = channels;
    const float* w = weights;
    do {
      float vacc0p0 = w[0];

      const float vi0 = *i0++;
      const float vk0 = w[1];
      vacc0p0 = math_muladd_f32(vi0, vk0, vacc0p0);

      const float vi1 = *i1++;
      const float vk1 = w[2];
      vacc0p0 = math_muladd_f32(vi1, vk1, vacc0p0);

      const float vi2 = *i2++;
      const float vk2 = w[3];
      vacc0p0 = math_muladd_f32(vi2, vk2, vacc0p0);

      const float vi3 = *i3++;
      const float vk3 = w[4];
      vacc0p0 = math_muladd_f32(vi3, vk3, vacc0p0);

      const float vi4 = *i4++;
      const float vk4 = w[5];
      vacc0p0 = math_muladd_f32(vi4, vk4, vacc0p0);

      const float vi5 = *i5++;
      const float vk5 = w[6];
      vacc0p0 = math_muladd_f32(vi5, vk5, vacc0p0);

      const float vi6 = *i6++;
      const float vk6 = w[7];
      vacc0p0 = math_muladd_f32(vi6, vk6, vacc0p0);

      const float vi7 = *i7++;
      const float vk7 = w[8];
      vacc0p0 = math_muladd_f32(vi7, vk7, vacc0p0);

      const float vi8 = *i8++;
      const float vk8 = w[9];
      vacc0p0 = math_muladd_f32(vi8, vk8, vacc0p0);

      const float vi9 = *i9++;
      const float vk9 = w[10];
      vacc0p0 = math_muladd_f32(vi9, vk9, vacc0p0);

      const float vi10 = *i10++;
      const float vk10 = w[11];
      vacc0p0 = math_muladd_f32(vi10, vk10, vacc0p0);

      const float vi11 = *i11++;
      const float vk11 = w[12];
      vacc0p0 = math_muladd_f32(vi11, vk11, vacc0p0);

      const float vi12 = *i12++;
      const float vk12 = w[13];
      vacc0p0 = math_muladd_f32(vi12, vk12, vacc0p0);

      const float vi13 = *i13++;
      const float vk13 = w[14];
      vacc0p0 = math_muladd_f32(vi13, vk13, vacc0p0);

      const float vi14 = *i14++;
      const float vk14 = w[15];
      vacc0p0 = math_muladd_f32(vi14, vk14, vacc0p0);

      const float vi15 = *i15++;
      const float vk15 = w[16];
      vacc0p0 = math_muladd_f32(vi15, vk15, vacc0p0);

      const float vi16 = *i16++;
      const float vk16 = w[17];
      vacc0p0 = math_muladd_f32(vi16, vk16, vacc0p0);

      const float vi17 = *i17++;
      const float vk17 = w[18];
      vacc0p0 = math_muladd_f32(vi17, vk17, vacc0p0);

      const float vi18 = *i18++;
      const float vk18 = w[19];
      vacc0p0 = math_muladd_f32(vi18, vk18, vacc0p0);

      const float vi19 = *i19++;
      const float vk19 = w[20];
      vacc0p0 = math_muladd_f32(vi19, vk19, vacc0p0);

      const float vi20 = *i20++;
      const float vk20 = w[21];
      vacc0p0 = math_muladd_f32(vi20, vk20, vacc0p0);

      const float vi21 = *i21++;
      const float vk21 = w[22];
      vacc0p0 = math_muladd_f32(vi21, vk21, vacc0p0);

      const float vi22 = *i22++;
      const float vk22 = w[23];
      vacc0p0 = math_muladd_f32(vi22, vk22, vacc0p0);

      const float vi23 = *i23++;
      const float vk23 = w[24];
      vacc0p0 = math_muladd_f32(vi23, vk23, vacc0p0);

      const float vi24 = *i24++;
      const float vk24 = w[25];
      vacc0p0 = math_muladd_f32(vi24, vk24, vacc0p0);

      w += 26;


      *output++ = vacc0p0;
    } while (--c != 0);

    output = (float*) ((uintptr_t) output + output_increment);
  } while (--output_width != 0);
}