rust-lang/rust
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Convert _mm256_insert_epi32 to const generics

Browse Source
This commit is contained in:
Rémy Rakic
2021-03-01 21:32:44 +01:00
committed by Amanieu d'Antras
parent a1ef46850a
commit e836a49019
2 changed files with 8 additions and 13 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

@@ -1527,16 +1527,11 @@ pub unsafe fn _mm256_insert_epi16<const INDEX: i32>(a: __m256i, i: i16) -> __m25
#[inline] #[inline]
#[target_feature(enable = "avx")] #[target_feature(enable = "avx")]
// This intrinsic has no corresponding instruction. // This intrinsic has no corresponding instruction.
#[rustc_args_required_const(2)] #[rustc_legacy_const_generics(2)]
#[stable(feature = "simd_x86", since = "1.27.0")] #[stable(feature = "simd_x86", since = "1.27.0")]
pub unsafe fn _mm256_insert_epi32(a: __m256i, i: i32, index: i32) -> __m256i { pub unsafe fn _mm256_insert_epi32<const INDEX: i32>(a: __m256i, i: i32) -> __m256i {
let a = a.as_i32x8(); static_assert_imm3!(INDEX);
macro_rules! call { transmute(simd_insert(a.as_i32x8(), INDEX as u32, i))
($index:expr) => {
simd_insert(a, $index, i)
};
}
transmute(constify_imm3!(index, call))
} }
/// Loads 256-bits (composed of 4 packed double-precision (64-bit) /// Loads 256-bits (composed of 4 packed double-precision (64-bit)
@@ -3990,7 +3985,7 @@ mod tests {
#[simd_test(enable = "avx")] #[simd_test(enable = "avx")]
unsafe fn test_mm256_insert_epi32() { unsafe fn test_mm256_insert_epi32() {
let a = _mm256_setr_epi32(1, 2, 3, 4, 5, 6, 7, 8); let a = _mm256_setr_epi32(1, 2, 3, 4, 5, 6, 7, 8);
let r = _mm256_insert_epi32(a, 0, 7); let r = _mm256_insert_epi32::<7>(a, 0);
let e = _mm256_setr_epi32(1, 2, 3, 4, 5, 6, 7, 0); let e = _mm256_setr_epi32(1, 2, 3, 4, 5, 6, 7, 0);
assert_eq_m256i(r, e); assert_eq_m256i(r, e);
} }

6
library/stdarch/crates/core_arch/src/x86/avx2.rs
View File

@@ -4552,7 +4552,7 @@ mod tests {
let b = _mm256_setr_epi32(7, 6, 2, 4, 3, 2, 1, 0); let b = _mm256_setr_epi32(7, 6, 2, 4, 3, 2, 1, 0);
let r = _mm256_cmpeq_epi32(a, b); let r = _mm256_cmpeq_epi32(a, b);
let e = _mm256_set1_epi32(0); let e = _mm256_set1_epi32(0);
let e = _mm256_insert_epi32(e, !0, 2); let e = _mm256_insert_epi32::<2>(e, !0);
assert_eq_m256i(r, e); assert_eq_m256i(r, e);
} }
@@ -4582,10 +4582,10 @@ mod tests {
#[simd_test(enable = "avx2")] #[simd_test(enable = "avx2")]
unsafe fn test_mm256_cmpgt_epi32() { unsafe fn test_mm256_cmpgt_epi32() {
let a = _mm256_insert_epi32(_mm256_set1_epi32(0), 5, 0); let a = _mm256_insert_epi32::<0>(_mm256_set1_epi32(0), 5);
let b = _mm256_set1_epi32(0); let b = _mm256_set1_epi32(0);
let r = _mm256_cmpgt_epi32(a, b); let r = _mm256_cmpgt_epi32(a, b);
assert_eq_m256i(r, _mm256_insert_epi32(_mm256_set1_epi32(0), !0, 0)); assert_eq_m256i(r, _mm256_insert_epi32::<0>(_mm256_set1_epi32(0), !0));
} }
#[simd_test(enable = "avx2")] #[simd_test(enable = "avx2")]