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

convert _mm_i32gather_epi32 to const generics

Browse source
This commit is contained in:
Rémy Rakic 2021-03-06 00:35:47 +01:00 • committed by Amanieu d'Antras
parent d4558e1b30
commit 48c7ba0886
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

@ -1085,20 +1085,19 @@ pub unsafe fn _mm256_hsubs_epi16(a: __m256i, b: __m256i) -> __m256i {
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_i32gather_epi32)
#[inline]
#[target_feature(enable = "avx2")]
#[cfg_attr(test, assert_instr(vpgatherdd, scale = 1))]
#[rustc_args_required_const(2)]
#[cfg_attr(test, assert_instr(vpgatherdd, SCALE = 1))]
#[rustc_legacy_const_generics(2)]
#[stable(feature = "simd_x86", since = "1.27.0")]
pub unsafe fn _mm_i32gather_epi32(slice: *const i32, offsets: __m128i, scale: i32) -> __m128i {
pub unsafe fn _mm_i32gather_epi32<const SCALE: i32>(
slice: *const i32,
offsets: __m128i,
) -> __m128i {
static_assert_imm8_scale!(SCALE);
let zero = _mm_setzero_si128().as_i32x4();
let neg_one = _mm_set1_epi32(-1).as_i32x4();
let offsets = offsets.as_i32x4();
let slice = slice as *const i8;
macro_rules! call {
($imm8:expr) => {
pgatherdd(zero, slice, offsets, neg_one, $imm8)
};
}
let r = constify_imm8_gather!(scale, call);
let r = pgatherdd(zero, slice, offsets, neg_one, SCALE as i8);
transmute(r)
}
@ -5573,7 +5572,7 @@ mod tests {
arr[i as usize] = i;
}
// A multiplier of 4 is word-addressing
let r = _mm_i32gather_epi32(arr.as_ptr(), _mm_setr_epi32(0, 16, 32, 48), 4);
let r = _mm_i32gather_epi32::<4>(arr.as_ptr(), _mm_setr_epi32(0, 16, 32, 48));
assert_eq_m128i(r, _mm_setr_epi32(0, 16, 32, 48));
}