1 // Auto-generated file. Do not edit!
2 // Template: src/f32-vbinary/vopc-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_vrdivc_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_vrdivc_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 const float vb = *b;
34 for (; n >= 4 * sizeof(float); n -= 4 * sizeof(float)) {
35 const float va0 = a[0];
36 const float va1 = a[1];
37 const float va2 = a[2];
38 const float va3 = a[3];
39 a += 4;
40
41 float vy0 = vb / va0;
42 float vy1 = vb / va1;
43 float vy2 = vb / va2;
44 float vy3 = vb / va3;
45
46
47 vy0 = math_max_f32(vy0, vy_min);
48 vy1 = math_max_f32(vy1, vy_min);
49 vy2 = math_max_f32(vy2, vy_min);
50 vy3 = math_max_f32(vy3, vy_min);
51
52 vy0 = math_min_f32(vy0, vy_max);
53 vy1 = math_min_f32(vy1, vy_max);
54 vy2 = math_min_f32(vy2, vy_max);
55 vy3 = math_min_f32(vy3, vy_max);
56
57 y[0] = vy0;
58 y[1] = vy1;
59 y[2] = vy2;
60 y[3] = vy3;
61 y += 4;
62 }
63 if XNN_UNLIKELY(n != 0) {
64 do {
65 const float va = *a++;
66 float vy = vb / va;
67 vy = math_max_f32(vy, vy_min);
68 vy = math_min_f32(vy, vy_max);
69 *y++ = vy;
70 n -= sizeof(float);
71 } while (n != 0);
72 }
73 }
74