1 // Auto-generated file. Do not edit!
2 //   Template: src/f32-vbinary/vop-scalar.c.in
3 //   Generator: tools/xngen
4 //
5 // Copyright 2019 Google LLC
6 //
7 // This source code is licensed under the BSD-style license found in the
8 // LICENSE file in the root directory of this source tree.
9 
10 #include <assert.h>
11 
12 #include <xnnpack/common.h>
13 #include <xnnpack/math.h>
14 #include <xnnpack/vbinary.h>
15 
16 
xnn_f32_vadd_minmax_ukernel__scalar_x4(size_t n,const float * a,const float * b,float * y,const union xnn_f32_minmax_params params[restrict XNN_MIN_ELEMENTS (1)])17 void xnn_f32_vadd_minmax_ukernel__scalar_x4(
18     size_t n,
19     const float* a,
20     const float* b,
21     float* y,
22     const union xnn_f32_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
23 {
24   assert(n != 0);
25   assert(n % sizeof(float) == 0);
26   assert(a != NULL);
27   assert(b != NULL);
28   assert(y != NULL);
29 
30   const float vy_min = params->scalar.min;
31   const float vy_max = params->scalar.max;
32 
33   for (; n >= 4 * sizeof(float); n -= 4 * sizeof(float)) {
34     const float va0 = a[0];
35     const float va1 = a[1];
36     const float va2 = a[2];
37     const float va3 = a[3];
38     a += 4;
39 
40     const float vb0 = b[0];
41     const float vb1 = b[1];
42     const float vb2 = b[2];
43     const float vb3 = b[3];
44     b += 4;
45 
46     float vy0 = va0 + vb0;
47     float vy1 = va1 + vb1;
48     float vy2 = va2 + vb2;
49     float vy3 = va3 + vb3;
50 
51 
52     vy0 = math_max_f32(vy0, vy_min);
53     vy1 = math_max_f32(vy1, vy_min);
54     vy2 = math_max_f32(vy2, vy_min);
55     vy3 = math_max_f32(vy3, vy_min);
56 
57     vy0 = math_min_f32(vy0, vy_max);
58     vy1 = math_min_f32(vy1, vy_max);
59     vy2 = math_min_f32(vy2, vy_max);
60     vy3 = math_min_f32(vy3, vy_max);
61 
62     y[0] = vy0;
63     y[1] = vy1;
64     y[2] = vy2;
65     y[3] = vy3;
66     y += 4;
67   }
68   if XNN_UNLIKELY(n != 0) {
69     do {
70       const float va = *a++;
71       const float vb = *b++;
72       float vy = va + vb;
73       vy = math_max_f32(vy, vy_min);
74       vy = math_min_f32(vy, vy_max);
75       *y++ = vy;
76       n -= sizeof(float);
77     } while (n != 0);
78   }
79 }
80