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

use stdsimd::simd::*;
use iters::{IntoPackedRefIterator, IntoPackedRefMutIterator, PackedIter};

#[cfg(all(target_feature = "avx512"))]
pub const SIMD_SIZE: usize = 512;

#[cfg(all(target_feature = "avx2", not(target_feature="avx512")))]
pub const SIMD_SIZE: usize = 256;

#[cfg(all(target_feature = "sse", not(target_feature="avx2")))]
pub const SIMD_SIZE: usize = 128;

#[cfg(all(target_arch = "x86", not(target_feature = "sse")))]
compile_error!("Your CPU is ancient! We don't support MMX, sorry!");

#[cfg(all(not(target_arch = "x86"), not(target_arch = "x86_64")))]
compile_error!("Support for non-x86 platforms is forthcoming - See the stdsimd issues tracker for more details.");

// A SIMD vector containing T.
pub trait Packed<T : Packable> {
    const WIDTH: usize;
    const ELEMENT_SIZE: usize = <T as Packable>::SIZE;
    const WIDTH_BYTES: usize = Self::WIDTH * Self::ELEMENT_SIZE / 8;

    #[inline(always)]
    fn width(&self) -> usize {
        Self::WIDTH
    }

    #[inline(always)]
    fn width_bytes(&self) -> usize {
        Self::WIDTH_BYTES
    }

    #[inline(always)]
    fn element_size(&self) -> usize {
        Self::ELEMENT_SIZE
    }

    fn load(data: &[T], offset: usize) -> Self;
    fn store(self, data: &mut [T], offset: usize);
    fn splat(data: T) -> Self;
}

// A type which may be packed into a SIMD vector
pub trait Packable where Self : Sized + Copy {
    const SIZE: usize;
    type Vector : Packed<Self>;
}

macro_rules! impl_packed {
    ($el:ty, $pvec:tt, $vec:tt, $sz:expr, $feat:expr, $nfeat:expr) => (
        #[allow(non_camel_case_types)]
        #[cfg(all(target_feature = $feat, not(target_feature = $nfeat)))]
        pub type $pvec = $vec;

        #[cfg(all(target_feature = $feat, not(target_feature = $nfeat)))]
        impl Packable for $el {
            const SIZE: usize = SIMD_SIZE / $sz;
            type Vector = $vec;
        }

        #[cfg(all(target_feature = $feat, not(target_feature = $nfeat)))]
        impl Packed<$el> for $vec {
            const WIDTH: usize = $sz;

            #[inline(always)]
            fn load(data: &[$el], offset: usize) -> $vec {
                $vec::load(data, offset)
            }
            #[inline(always)]
            fn store(self, data: &mut [$el], offset: usize) {
                $vec::store(self, data, offset);
            }
            #[inline(always)]
            fn splat(data: $el) -> Self {
                $vec::splat(data)
            }
        }

        #[cfg(all(target_feature = $feat, not(target_feature = $nfeat)))]
        impl<'a> IntoPackedRefIterator<'a> for &'a [$el] {
            type Iter = PackedIter<'a, $el>;

            #[inline(always)]
            fn simd_iter(&'a self) -> Self::Iter {
                PackedIter {
                    data: self,
                    position: 0,
                }
            }
        }

        #[cfg(all(target_feature = $feat, not(target_feature = $nfeat)))]
        impl<'a> IntoPackedRefMutIterator<'a> for &'a mut [$el] {
            type Iter = PackedIter<'a, $el>;

            #[inline(always)]
            fn simd_iter_mut(&'a mut self) -> Self::Iter {
                PackedIter {
                    data: self,
                    position: 0,
                }
            }
        }
    );
}

impl_packed!(u8, u8s, u8x64, 64, "avx512", "avx1024");
impl_packed!(u8, u8s, u8x32, 32, "avx2", "avx512");
impl_packed!(u8, u8s, u8x16, 16, "sse", "avx2");

impl_packed!(i8, i8s, i8x64, 64, "avx512", "avx1024");
impl_packed!(i8, i8s, i8x32, 32, "avx2", "avx512");
impl_packed!(i8, i8s, i8x16, 16, "sse", "avx2");

impl_packed!(u16, u16s, u16x32, 32, "avx512", "avx1024");
impl_packed!(u16, u16s, u16x16, 16, "avx2", "avx512");
impl_packed!(u16, u16s, u16x8, 8, "sse", "avx2");

impl_packed!(i16, i16s, i16x32, 32, "avx512", "avx1024");
impl_packed!(i16, i16s, i16x16, 16, "avx2", "avx512");
impl_packed!(i16, i16s, i16x8, 8, "sse", "avx2");

impl_packed!(u32, u32s, u32x16, 16, "avx512", "avx1024");
impl_packed!(u32, u32s, u32x8, 8, "avx2", "avx512");
impl_packed!(u32, u32s, u32x4, 4, "sse", "avx2");

impl_packed!(i32, i32s, i32x16, 16, "avx512", "avx1024");
impl_packed!(i32, i32s, i32x8, 8, "avx2", "avx512");
impl_packed!(i32, i32s, i32x4, 4, "sse", "avx2");

impl_packed!(f32, f32s, f32x16, 16, "avx512", "avx1024");
impl_packed!(f32, f32s, f32x8, 8, "avx2", "avx512");
impl_packed!(f32, f32s, f32x4, 4, "sse", "avx2");

impl_packed!(u64, u64s, u64x8, 8, "avx512", "avx1024");
impl_packed!(u64, u64s, u64x4, 4, "avx2", "avx512");
impl_packed!(u64, u64s, u64x2, 2, "sse", "avx2");

impl_packed!(i64, i64s, i64x8, 8, "avx512", "avx1024");
impl_packed!(i64, i64s, i64x4, 4, "avx2", "avx512");
impl_packed!(i64, i64s, i64x2, 2, "sse", "avx2");

impl_packed!(f64, f64s, f64x8, 8, "avx512", "avx1024");
impl_packed!(f64, f64s, f64x4, 4, "avx2", "avx512");
impl_packed!(f64, f64s, f64x2, 2, "sse", "avx2");