hermes-simd 0.6.0

High-performance, zero-overhead SIMD abstraction library
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

//! Generic runtime-dispatch sum kernel.

use hermes_simd_core::{
    align::Unaligned, arch::SimdArch, execution::Unmasked, kernel::SimdKernel, scalar::Scalar,
    view::SimdView,
};
use hermes_simd_macros::runtime_dispatch;

#[runtime_dispatch(avx512f, avx2, neon, scalar)]
pub(super) fn dispatch_sum_kernel<T, A>(data: &[T]) -> T
where
    T: Scalar,
    A: SimdArch + SimdKernel<T>,
{
    match SimdView::<T, A, Unaligned, Unmasked, &[T]>::new(data) {
        Some(v) => v.sum(),
        None => unsafe { core::hint::unreachable_unchecked() },
    }
}