1 // Auto-generated file. Do not edit!
2 // Template: src/f32-vunary/neon.c.in
3 // Generator: tools/xngen
4 //
5 // Copyright 2020 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 <arm_neon.h>
13
14 #include <xnnpack/common.h>
15 #include <xnnpack/vunary.h>
16
17
xnn_f32_vneg_ukernel__neon_x8(size_t n,const float * x,float * y,const union xnn_f32_neg_params params[restrict XNN_MIN_ELEMENTS (1)])18 void xnn_f32_vneg_ukernel__neon_x8(
19 size_t n,
20 const float* x,
21 float* y,
22 const union xnn_f32_neg_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_DISABLE_TSAN
23 {
24 assert(n != 0);
25 assert(n % sizeof(float) == 0);
26 assert(x != NULL);
27 assert(y != NULL);
28
29 for (; n >= 8 * sizeof(float); n -= 8 * sizeof(float)) {
30 const float32x4_t vx0123 = vld1q_f32(x); x += 4;
31 const float32x4_t vx4567 = vld1q_f32(x); x += 4;
32
33 const float32x4_t vy0123 = vnegq_f32(vx0123);
34 const float32x4_t vy4567 = vnegq_f32(vx4567);
35
36 vst1q_f32(y, vy0123); y += 4;
37 vst1q_f32(y, vy4567); y += 4;
38 }
39 for (; n >= 4 * sizeof(float); n -= 4 * sizeof(float)) {
40 const float32x4_t vx = vld1q_f32(x); x += 4;
41 const float32x4_t vy = vnegq_f32(vx);
42 vst1q_f32(y, vy); y += 4;
43 }
44 if XNN_UNLIKELY(n != 0) {
45 const float32x4_t vx = vld1q_f32(x);
46 const float32x4_t vy = vnegq_f32(vx);
47 float32x2_t vy_lo = vget_low_f32(vy);
48 if (n & (2 * sizeof(float))) {
49 vst1_f32(y, vy_lo); y += 2;
50 vy_lo = vget_high_f32(vy);
51 }
52 if (n & (1 * sizeof(float))) {
53 vst1_lane_f32(y, vy_lo, 0);
54 }
55 }
56 }
57