Function safe_arch::load_masked_m256

#[must_use]pub fn load_masked_m256(a: &m256, mask: m256i) -> m256

This is supported with target feature avx only.

Load data from memory into a register according to a mask.

When the high bit of a mask lane isn't set the loaded lane will be zero.

let a = m256::from([8.0, 17.0, 16.0, 20.0, 80.0, 1.0, 2.0, 3.0]);
let b =
  load_masked_m256(&a, m256i::from([0, -1, -1, 0, -1, -1, 0, 0])).to_array();
assert_eq!(b, [0.0, 17.0, 16.0, 0.0, 80.0, 1.0, 0.0, 0.0]);