thermite 0.2.1

High-performance, generic, ISA-portable SIMD library with a policy-configurable transcendental math library
Documentation 
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

use generic_array::{GenericArray, typenum};

use crate::{
    Vector,
    divider::Denominator,
    register::{Element, IntegerRegister, Lanes, Register},
    vector::{GenericVector as _, VectorWithRegister as _},
};

/// Precomputed multipliers and shifts for branchless vectorized division
pub struct VectorDivider<R: Register> {
    pub multipliers: Vector<R>,
    pub shifts: Vector<R>,
}

impl<R: Register> Clone for VectorDivider<R> {
    #[inline(always)]
    fn clone(&self) -> Self {
        *self
    }
}

impl<R: Register> Copy for VectorDivider<R> {}

impl<R: IntegerRegister> VectorDivider<R>
where
    R::Element: Denominator,
{
    #[inline]
    pub fn new(divisor: Vector<R>) -> Self {
        Self::try_new(divisor).unwrap()
    }

    #[inline]
    pub fn try_new(divisor: Vector<R>) -> Result<Self, super::UnsupportedDivisor> {
        let mut multipliers = GenericArray::default();
        let mut shifts = GenericArray::default();

        for ((m, s), d) in multipliers.iter_mut().zip(shifts.iter_mut()).zip(divisor.as_slice()) {
            let divisor = d.try_to_branchfree_divider()?;

            *m = divisor.multiplier();
            *s = Element::from_i8(divisor.shift() as i8);
        }

        Ok(VectorDivider {
            multipliers: Vector(R::new(multipliers)),
            shifts: Vector(R::new(shifts)),
        })
    }
}