• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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_relu_ukernel__scalar_x4(size_t n,const float * a,const float * b,float * y,const union xnn_f32_relu_params params[restrict XNN_MIN_ELEMENTS (1)])17 void xnn_f32_vrdivc_relu_ukernel__scalar_x4(
18     size_t n,
19     const float* a,
20     const float* b,
21     float* y,
22     const union xnn_f32_relu_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 
31   const float vb = *b;
32   for (; n >= 4 * sizeof(float); n -= 4 * sizeof(float)) {
33     const float va0 = a[0];
34     const float va1 = a[1];
35     const float va2 = a[2];
36     const float va3 = a[3];
37     a += 4;
38 
39     float vy0 = vb / va0;
40     float vy1 = vb / va1;
41     float vy2 = vb / va2;
42     float vy3 = vb / va3;
43 
44 
45     vy0 = math_max_f32(vy0, 0.0f);
46     vy1 = math_max_f32(vy1, 0.0f);
47     vy2 = math_max_f32(vy2, 0.0f);
48     vy3 = math_max_f32(vy3, 0.0f);
49 
50     y[0] = vy0;
51     y[1] = vy1;
52     y[2] = vy2;
53     y[3] = vy3;
54     y += 4;
55   }
56   if XNN_UNLIKELY(n != 0) {
57     do {
58       const float va = *a++;
59       float vy = vb / va;
60       vy = math_max_f32(vy, 0.0f);
61       *y++ = vy;
62       n -= sizeof(float);
63     } while (n != 0);
64   }
65 }
66