fearless_simd 0.4.0

Safer and easier 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
41
42
43
44
45
46
47
48

// Copyright 2024 the Fearless_SIMD Authors
// SPDX-License-Identifier: Apache-2.0 OR MIT

//! Access to SSE3 intrinsics.

use crate::impl_macros::delegate;
#[cfg(target_arch = "x86")]
use core::arch::x86 as arch;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64 as arch;

use arch::*;

/// A token for SSE3 intrinsics on `x86` and `x86_64`.
#[derive(Clone, Copy, Debug)]
pub struct Sse3 {
    _private: (),
}

#[expect(
    clippy::missing_safety_doc,
    reason = "TODO: https://github.com/linebender/fearless_simd/issues/40"
)]
impl Sse3 {
    /// Create a SIMD token.
    ///
    /// # Safety
    ///
    /// The required CPU features must be available.
    #[inline]
    pub const unsafe fn new_unchecked() -> Self {
        Self { _private: () }
    }

    delegate! { arch:
        fn _mm_addsub_ps(a: __m128, b: __m128) -> __m128;
        fn _mm_addsub_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_hadd_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_hadd_ps(a: __m128, b: __m128) -> __m128;
        fn _mm_hsub_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_hsub_ps(a: __m128, b: __m128) -> __m128;
        unsafe fn _mm_lddqu_si128(mem_addr: *const __m128i) -> __m128i;
        fn _mm_movedup_pd(a: __m128d) -> __m128d;
        unsafe fn _mm_loaddup_pd(mem_addr: *const f64) -> __m128d;
        fn _mm_movehdup_ps(a: __m128) -> __m128;
        fn _mm_moveldup_ps(a: __m128) -> __m128;
    }
}