1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
|
/*
* Copyright (c) 2015 Cisco and/or its affiliates.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef included_vector_neon_h
#define included_vector_neon_h
#include <arm_neon.h>
/* Arithmetic */
#define u16x8_sub_saturate(a,b) vsubq_u16(a,b)
#define i16x8_sub_saturate(a,b) vsubq_s16(a,b)
/* Converts all ones/zeros compare mask to bitmap. */
always_inline u32
u8x16_compare_byte_mask (u8x16 x)
{
uint8x16_t mask_shift =
{ -7, -6, -5, -4, -3, -2, -1, 0, -7, -6, -5, -4, -3, -2, -1, 0 };
uint8x16_t mask_and = vdupq_n_u8 (0x80);
x = vandq_u8 (x, mask_and);
x = vshlq_u8 (x, vreinterpretq_s8_u8 (mask_shift));
x = vpaddq_u8 (x, x);
x = vpaddq_u8 (x, x);
x = vpaddq_u8 (x, x);
return vgetq_lane_u8 (x, 0) | (vgetq_lane_u8 (x, 1) << 8);
}
always_inline u32
u16x8_zero_byte_mask (u16x8 input)
{
u8x16 vall_one = vdupq_n_u8 (0x0);
u8x16 res_values = { 0x01, 0x02, 0x04, 0x08,
0x10, 0x20, 0x40, 0x80,
0x01, 0x02, 0x04, 0x08,
0x10, 0x20, 0x40, 0x80
};
/* input --> [0x80, 0x40, 0x01, 0xf0, ... ] */
u8x16 test_result =
vreinterpretq_u8_u16 (vceqq_u16 (input, vreinterpretq_u16_u8 (vall_one)));
u8x16 before_merge = vminq_u8 (test_result, res_values);
/*before_merge--> [0x80, 0x00, 0x00, 0x10, ... ] */
/* u8x16 --> [a,b,c,d, e,f,g,h, i,j,k,l, m,n,o,p] */
/* pair add until we have 2 uint64_t */
u16x8 merge1 = vpaddlq_u8 (before_merge);
/* u16x8--> [a+b,c+d, e+f,g+h, i+j,k+l, m+n,o+p] */
u32x4 merge2 = vpaddlq_u16 (merge1);
/* u32x4--> [a+b+c+d, e+f+g+h, i+j+k+l, m+n+o+p] */
u64x2 merge3 = vpaddlq_u32 (merge2);
/* u64x2--> [a+b+c+d+e+f+g+h, i+j+k+l+m+n+o+p] */
return (u32) (vgetq_lane_u64 (merge3, 1) << 8) + vgetq_lane_u64 (merge3, 0);
}
always_inline u32
u8x16_zero_byte_mask (u8x16 input)
{
return u16x8_zero_byte_mask ((u16x8) input);
}
always_inline u32
u32x4_zero_byte_mask (u32x4 input)
{
return u16x8_zero_byte_mask ((u16x8) input);
}
always_inline u32
u64x2_zero_byte_mask (u64x2 input)
{
return u16x8_zero_byte_mask ((u16x8) input);
}
/* *INDENT-OFF* */
#define foreach_neon_vec128i \
_(i,8,16,s8) _(i,16,8,s16) _(i,32,4,s32) _(i,64,2,s64)
#define foreach_neon_vec128u \
_(u,8,16,u8) _(u,16,8,u16) _(u,32,4,u32) _(u,64,2,u64)
#define foreach_neon_vec128f \
_(f,32,4,f32) _(f,64,2,f64)
#define _(t, s, c, i) \
static_always_inline t##s##x##c \
t##s##x##c##_splat (t##s x) \
{ return (t##s##x##c) vdupq_n_##i (x); } \
\
static_always_inline t##s##x##c \
t##s##x##c##_load_unaligned (void *p) \
{ return (t##s##x##c) vld1q_##i (p); } \
\
static_always_inline void \
t##s##x##c##_store_unaligned (t##s##x##c v, void *p) \
{ vst1q_##i (p, v); } \
\
static_always_inline int \
t##s##x##c##_is_all_zero (t##s##x##c x) \
{ return !(vaddvq_##i (x)); } \
\
static_always_inline int \
t##s##x##c##_is_equal (t##s##x##c a, t##s##x##c b) \
{ return t##s##x##c##_is_all_zero (a ^ b); } \
\
static_always_inline int \
t##s##x##c##_is_all_equal (t##s##x##c v, t##s x) \
{ return t##s##x##c##_is_equal (v, t##s##x##c##_splat (x)); }; \
foreach_neon_vec128i foreach_neon_vec128u
#undef _
/* *INDENT-ON* */
static_always_inline u16x8
u16x8_byte_swap (u16x8 v)
{
return (u16x8) vrev16q_u8 ((u8x16) v);
}
static_always_inline u8x16
u8x16_shuffle (u8x16 v, u8x16 m)
{
return (u8x16) vqtbl1q_u8 (v, m);
}
static_always_inline u32x4
u32x4_hadd (u32x4 v1, u32x4 v2)
{
return (u32x4) vpaddq_u32 (v1, v2);
}
static_always_inline u64x2
u32x4_extend_to_u64x2 (u32x4 v)
{
return vmovl_u32 (vget_low_u32 (v));
}
static_always_inline u64x2
u32x4_extend_to_u64x2_high (u32x4 v)
{
return vmovl_high_u32 (v);
}
#define CLIB_HAVE_VEC128_UNALIGNED_LOAD_STORE
#define CLIB_VEC128_SPLAT_DEFINED
#endif /* included_vector_neon_h */
/*
* fd.io coding-style-patch-verification: ON
*
* Local Variables:
* eval: (c-set-style "gnu")
* End:
*/
|
