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

Convert _mm256_insert_epi16 to const generics

Browse source
This commit is contained in:
Rémy Rakic 2021-03-01 21:29:50 +01:00 • committed by Amanieu d'Antras
parent 7d269db1c0
commit a1ef46850a
2 changed files with 13 additions and 18 deletions
Download patch file Download diff file Expand all files Collapse all files

15
library/stdarch/crates/core_arch/src/x86/avx.rs
View file

@ -1513,16 +1513,11 @@ pub unsafe fn _mm256_insert_epi8(a: __m256i, i: i8, index: i32) -> __m256i {
#[inline]
#[target_feature(enable = "avx")]
// This intrinsic has no corresponding instruction.
#[rustc_args_required_const(2)]
#[rustc_legacy_const_generics(2)]
#[stable(feature = "simd_x86", since = "1.27.0")]
pub unsafe fn _mm256_insert_epi16(a: __m256i, i: i16, index: i32) -> __m256i {
let a = a.as_i16x16();
macro_rules! call {
($index:expr) => {
simd_insert(a, $index, i)
};
}
transmute(constify_imm4!((index & 15), call))
pub unsafe fn _mm256_insert_epi16<const INDEX: i32>(a: __m256i, i: i16) -> __m256i {
static_assert_imm4!(INDEX);
transmute(simd_insert(a.as_i16x16(), INDEX as u32, i))
}
/// Copies `a` to result, and inserts the 32-bit integer `i` into result
@ -3983,7 +3978,7 @@ mod tests {
0, 1, 2, 3, 4, 5, 6, 7,
8, 9, 10, 11, 12, 13, 14, 15,
);
let r = _mm256_insert_epi16(a, 0, 15);
let r = _mm256_insert_epi16::<15>(a, 0);
#[rustfmt::skip]
let e = _mm256_setr_epi16(
0, 1, 2, 3, 4, 5, 6, 7,

16
library/stdarch/crates/core_arch/src/x86/avx2.rs
View file

@ -4543,7 +4543,7 @@ mod tests {
7, 6, 5, 4, 3, 2, 1, 0,
);
let r = _mm256_cmpeq_epi16(a, b);
assert_eq_m256i(r, _mm256_insert_epi16(_mm256_set1_epi16(0), !0, 2));
assert_eq_m256i(r, _mm256_insert_epi16::<2>(_mm256_set1_epi16(0), !0));
}
#[simd_test(enable = "avx2")]
@ -4574,10 +4574,10 @@ mod tests {
#[simd_test(enable = "avx2")]
unsafe fn test_mm256_cmpgt_epi16() {
let a = _mm256_insert_epi16(_mm256_set1_epi16(0), 5, 0);
let a = _mm256_insert_epi16::<0>(_mm256_set1_epi16(0), 5);
let b = _mm256_set1_epi16(0);
let r = _mm256_cmpgt_epi16(a, b);
assert_eq_m256i(r, _mm256_insert_epi16(_mm256_set1_epi16(0), !0, 0));
assert_eq_m256i(r, _mm256_insert_epi16::<0>(_mm256_set1_epi16(0), !0));
}
#[simd_test(enable = "avx2")]
@ -4741,8 +4741,8 @@ mod tests {
#[simd_test(enable = "avx2")]
unsafe fn test_mm256_hadds_epi16() {
let a = _mm256_set1_epi16(2);
let a = _mm256_insert_epi16(a, 0x7fff, 0);
let a = _mm256_insert_epi16(a, 1, 1);
let a = _mm256_insert_epi16::<0>(a, 0x7fff);
let a = _mm256_insert_epi16::<1>(a, 1);
let b = _mm256_set1_epi16(4);
let r = _mm256_hadds_epi16(a, b);
#[rustfmt::skip]
@ -4774,11 +4774,11 @@ mod tests {
#[simd_test(enable = "avx2")]
unsafe fn test_mm256_hsubs_epi16() {
let a = _mm256_set1_epi16(2);
let a = _mm256_insert_epi16(a, 0x7fff, 0);
let a = _mm256_insert_epi16(a, -1, 1);
let a = _mm256_insert_epi16::<0>(a, 0x7fff);
let a = _mm256_insert_epi16::<1>(a, -1);
let b = _mm256_set1_epi16(4);
let r = _mm256_hsubs_epi16(a, b);
let e = _mm256_insert_epi16(_mm256_set1_epi16(0), 0x7FFF, 0);
let e = _mm256_insert_epi16::<0>(_mm256_set1_epi16(0), 0x7FFF);
assert_eq_m256i(r, e);
}