1 // Auto-generated file. Do not edit!
2 // Template: src/f32-vbinary/vopc-wasmsimd.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 <wasm_simd128.h>
13
14 #include <xnnpack/common.h>
15 #include <xnnpack/vbinary.h>
16
17
xnn_f32_vsqrdiffc_ukernel__wasmsimd_x8(size_t n,const float * a,const float * b,float * y,const union xnn_f32_default_params params[restrict XNN_MIN_ELEMENTS (1)])18 void xnn_f32_vsqrdiffc_ukernel__wasmsimd_x8(
19 size_t n,
20 const float* a,
21 const float* b,
22 float* y,
23 const union xnn_f32_default_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_DISABLE_TSAN
24 {
25 assert(n != 0);
26 assert(n % sizeof(float) == 0);
27 assert(a != NULL);
28 assert(b != NULL);
29 assert(y != NULL);
30
31 const v128_t vb = wasm_v32x4_load_splat(b);
32 for (; n >= 8 * sizeof(float); n -= 8 * sizeof(float)) {
33 const v128_t va0123 = wasm_v128_load(a);
34 const v128_t va4567 = wasm_v128_load(a + 4);
35 a += 8;
36
37 v128_t vy0123 = wasm_f32x4_sub(va0123, vb);
38 v128_t vy4567 = wasm_f32x4_sub(va4567, vb);
39
40 vy0123 = wasm_f32x4_mul(vy0123, vy0123);
41 vy4567 = wasm_f32x4_mul(vy4567, vy4567);
42
43
44 wasm_v128_store(y, vy0123);
45 wasm_v128_store(y + 4, vy4567);
46 y += 8;
47 }
48 for (; n >= 4 * sizeof(float); n -= 4 * sizeof(float)) {
49 const v128_t va = wasm_v128_load(a);
50 a += 4;
51
52 v128_t vy = wasm_f32x4_sub(va, vb);
53 vy = wasm_f32x4_mul(vy, vy);
54
55
56 wasm_v128_store(y, vy);
57 y += 4;
58 }
59 if XNN_UNLIKELY(n != 0) {
60 const v128_t va = wasm_v128_load(a);
61
62 v128_t vy = wasm_f32x4_sub(va, vb);
63 vy = wasm_f32x4_mul(vy, vy);
64
65
66 if (n & (2 * sizeof(float))) {
67 *((double*) y) = wasm_f64x2_extract_lane(vy, 0);
68 vy = wasm_v32x4_shuffle(vy, vy, 2, 3, 2, 3);
69 y += 2;
70 }
71 if (n & (1 * sizeof(float))) {
72 *y = wasm_f32x4_extract_lane(vy, 0);
73 }
74 }
75 }
76