optimath 0.3.1

Linear Algebra using const generics for no_std and specialization to enable SIMD.
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

use crate::Vector;
use core::{
	arch::x86_64::{_mm256_add_ps, _mm256_loadu_ps, _mm256_store_ps},
	mem::size_of,
	ops::Add,
};

impl<'a, const N: usize> Add for &'a Vector<f32, N> {
	fn add(self, other: Self) -> Vector<f32, N> {
		let simd_width = 256 / 8;
		let simd_elements = simd_width / size_of::<f32>();

		assert_eq!(simd_elements, 8);

		let simd_self = self.inner.chunks_exact(simd_elements);
		let remainder_self = simd_self.remainder();

		let simd_other = other.inner.chunks_exact(simd_elements);
		let remainder_other = simd_other.remainder();

		let simd = simd_self
			.zip(simd_other)
			.map(|(s, o)| unsafe {
				let s = _mm256_loadu_ps(s.as_ptr());
				let o = _mm256_loadu_ps(o.as_ptr());
				let res = _mm256_add_ps(s, o);
				let mut dst = [0.; 8];
				_mm256_store_ps(dst.as_mut_ptr(), res);
				Vector { inner: dst }
			})
			.flatten();

		let remainder = remainder_self
			.iter()
			.zip(remainder_other)
			.map(|(s, o)| Add::add(s, o));

		simd.chain(remainder).collect()
	}
}