search

Lokathor / safe_arch

Exposes arch-specific intrinsics as safe function (via cfg).
https://docs.rs/safe_arch
Apache License 2.0
47 stars 8 forks source link

fma intrinsics #70

Closed Soveu closed 4 years ago

Soveu commented 4 years ago

Closes #47

Lokathor commented 4 years ago

This looks good, can you add the Intrinsic and Assembly entries to each doc comment block underneath the doctest like I'm starting to do with the avx2 content?

Here's a sample from that branch:

/// Lanewise `max(a, b)` with lanes as `i32`.
/// ```
/// # use safe_arch::*;
/// let a = m256i::from([0_i32, 1, 2, 3, 4, 5, 6, 7]);
/// let b = m256i::from([0_i32, 11, 2, -13, 4, 15, 6, -17]);
/// let c: [i32; 8] = max_i32_m256i(a, b).into();
/// assert_eq!(c, [0, 11, 2, 3, 4, 15, 6, 7]);
/// ```
/// * **Intrinsic:** [`_mm256_max_epi32`]
/// * **Assembly:** `vpmaxsd ymm, ymm, ymm`
#[must_use]
#[inline(always)]
#[cfg_attr(docs_rs, doc(cfg(target_feature = "avx2")))]
pub fn max_i32_m256i(a: m256i, b: m256i) -> m256i {
  m256i(unsafe { _mm256_max_epi32(a.0, b.0) })
}