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

use crate::simd::intrinsics::{
    simd_reduce_add_ordered, simd_reduce_and, simd_reduce_max, simd_reduce_min,
    simd_reduce_mul_ordered, simd_reduce_or, simd_reduce_xor,
};
use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount};
use core::ops::{BitAnd, BitOr, BitXor};

macro_rules! impl_integer_reductions {
    { $scalar:ty } => {
        impl<const LANES: usize> Simd<$scalar, LANES>
        where
            LaneCount<LANES>: SupportedLaneCount,
        {
            /// Horizontal wrapping add.  Returns the sum of the lanes of the vector, with wrapping addition.
            #[inline]
            pub fn horizontal_sum(self) -> $scalar {
                // Safety: `self` is an integer vector
                unsafe { simd_reduce_add_ordered(self, 0) }
            }

            /// Horizontal wrapping multiply.  Returns the product of the lanes of the vector, with wrapping multiplication.
            #[inline]
            pub fn horizontal_product(self) -> $scalar {
                // Safety: `self` is an integer vector
                unsafe { simd_reduce_mul_ordered(self, 1) }
            }

            /// Horizontal maximum.  Returns the maximum lane in the vector.
            #[inline]
            pub fn horizontal_max(self) -> $scalar {
                // Safety: `self` is an integer vector
                unsafe { simd_reduce_max(self) }
            }

            /// Horizontal minimum.  Returns the minimum lane in the vector.
            #[inline]
            pub fn horizontal_min(self) -> $scalar {
                // Safety: `self` is an integer vector
                unsafe { simd_reduce_min(self) }
            }
        }
    }
}

impl_integer_reductions! { i8 }
impl_integer_reductions! { i16 }
impl_integer_reductions! { i32 }
impl_integer_reductions! { i64 }
impl_integer_reductions! { isize }
impl_integer_reductions! { u8 }
impl_integer_reductions! { u16 }
impl_integer_reductions! { u32 }
impl_integer_reductions! { u64 }
impl_integer_reductions! { usize }

macro_rules! impl_float_reductions {
    { $scalar:ty } => {
        impl<const LANES: usize> Simd<$scalar, LANES>
        where
            LaneCount<LANES>: SupportedLaneCount,
        {

            /// Horizontal add.  Returns the sum of the lanes of the vector.
            #[inline]
            pub fn horizontal_sum(self) -> $scalar {
                // LLVM sum is inaccurate on i586
                if cfg!(all(target_arch = "x86", not(target_feature = "sse2"))) {
                    self.as_array().iter().sum()
                } else {
                    // Safety: `self` is a float vector
                    unsafe { simd_reduce_add_ordered(self, 0.) }
                }
            }

            /// Horizontal multiply.  Returns the product of the lanes of the vector.
            #[inline]
            pub fn horizontal_product(self) -> $scalar {
                // LLVM product is inaccurate on i586
                if cfg!(all(target_arch = "x86", not(target_feature = "sse2"))) {
                    self.as_array().iter().product()
                } else {
                    // Safety: `self` is a float vector
                    unsafe { simd_reduce_mul_ordered(self, 1.) }
                }
            }

            /// Horizontal maximum.  Returns the maximum lane in the vector.
            ///
            /// Returns values based on equality, so a vector containing both `0.` and `-0.` may
            /// return either.  This function will not return `NaN` unless all lanes are `NaN`.
            #[inline]
            pub fn horizontal_max(self) -> $scalar {
                // Safety: `self` is a float vector
                unsafe { simd_reduce_max(self) }
            }

            /// Horizontal minimum.  Returns the minimum lane in the vector.
            ///
            /// Returns values based on equality, so a vector containing both `0.` and `-0.` may
            /// return either.  This function will not return `NaN` unless all lanes are `NaN`.
            #[inline]
            pub fn horizontal_min(self) -> $scalar {
                // Safety: `self` is a float vector
                unsafe { simd_reduce_min(self) }
            }
        }
    }
}

impl_float_reductions! { f32 }
impl_float_reductions! { f64 }

impl<T, const LANES: usize> Simd<T, LANES>
where
    Self: BitAnd<Self, Output = Self>,
    T: SimdElement + BitAnd<T, Output = T>,
    LaneCount<LANES>: SupportedLaneCount,
{
    /// Horizontal bitwise "and".  Returns the cumulative bitwise "and" across the lanes of
    /// the vector.
    #[inline]
    pub fn horizontal_and(self) -> T {
        unsafe { simd_reduce_and(self) }
    }
}

impl<T, const LANES: usize> Simd<T, LANES>
where
    Self: BitOr<Self, Output = Self>,
    T: SimdElement + BitOr<T, Output = T>,
    LaneCount<LANES>: SupportedLaneCount,
{
    /// Horizontal bitwise "or".  Returns the cumulative bitwise "or" across the lanes of
    /// the vector.
    #[inline]
    pub fn horizontal_or(self) -> T {
        unsafe { simd_reduce_or(self) }
    }
}

impl<T, const LANES: usize> Simd<T, LANES>
where
    Self: BitXor<Self, Output = Self>,
    T: SimdElement + BitXor<T, Output = T>,
    LaneCount<LANES>: SupportedLaneCount,
{
    /// Horizontal bitwise "xor".  Returns the cumulative bitwise "xor" across the lanes of
    /// the vector.
    #[inline]
    pub fn horizontal_xor(self) -> T {
        unsafe { simd_reduce_xor(self) }
    }
}