user0/rust
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

convert _mm256_i32gather_epi64 to const generics

Browse source
This commit is contained in:
Rémy Rakic 2021-03-06 00:55:31 +01:00 • committed by Amanieu d'Antras
parent e1a667eda9
commit 771603ebf5
1 changed files with 9 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

19
library/stdarch/crates/core_arch/src/x86/avx2.rs
View file

@ -1316,20 +1316,19 @@ pub unsafe fn _mm_mask_i32gather_epi64<const SCALE: i32>(
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm256_i32gather_epi64)
#[inline]
#[target_feature(enable = "avx2")]
#[cfg_attr(test, assert_instr(vpgatherdq, scale = 1))]
#[rustc_args_required_const(2)]
#[cfg_attr(test, assert_instr(vpgatherdq, SCALE = 1))]
#[rustc_legacy_const_generics(2)]
#[stable(feature = "simd_x86", since = "1.27.0")]
pub unsafe fn _mm256_i32gather_epi64(slice: *const i64, offsets: __m128i, scale: i32) -> __m256i {
pub unsafe fn _mm256_i32gather_epi64<const SCALE: i32>(
slice: *const i64,
offsets: __m128i,
) -> __m256i {
static_assert_imm8_scale!(SCALE);
let zero = _mm256_setzero_si256().as_i64x4();
let neg_one = _mm256_set1_epi64x(-1).as_i64x4();
let offsets = offsets.as_i32x4();
let slice = slice as *const i8;
macro_rules! call {
($imm8:expr) => {
vpgatherdq(zero, slice, offsets, neg_one, $imm8)
};
}
let r = constify_imm8_gather!(scale, call);
let r = vpgatherdq(zero, slice, offsets, neg_one, SCALE as i8);
transmute(r)
}
@ -5685,7 +5684,7 @@ mod tests {
arr[i as usize] = i;
}
// A multiplier of 8 is word-addressing for i64s
let r = _mm256_i32gather_epi64(arr.as_ptr(), _mm_setr_epi32(0, 16, 32, 48), 8);
let r = _mm256_i32gather_epi64::<8>(arr.as_ptr(), _mm_setr_epi32(0, 16, 32, 48));
assert_eq_m256i(r, _mm256_setr_epi64x(0, 16, 32, 48));
}