1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
|
// run-pass
// only-x86_64
// compile-flags: -C opt-level=3
// Regression test for issue #79865.
// The assertion will fail when compiled with Rust 1.56..=1.59
// due to an LLVM miscompilation.
use std::arch::x86_64::*;
fn main() {
if is_x86_feature_detected!("avx") {
let res: [f64; 4] = unsafe { std::mem::transmute::<_, _>(first()) };
assert_eq!(res, [22.0, 44.0, 66.0, 88.0]);
}
}
#[target_feature(enable = "avx")]
unsafe fn first() -> __m256d {
second()
}
unsafe fn second() -> __m256d {
let v0 = _mm256_setr_pd(1.0, 2.0, 3.0, 4.0);
let v1 = _mm256_setr_pd(10.0, 20.0, 30.0, 40.0);
// needs to be called twice to hit the miscompilation
let (add, _) = add_sub(v0, v1);
let (add, _) = add_sub(add, add);
add
}
#[inline(never)] // needed to hit the miscompilation
unsafe fn add_sub(v1: __m256d, v0: __m256d) -> (__m256d, __m256d) {
let add = _mm256_add_pd(v0, v1);
let sub = _mm256_sub_pd(v0, v1);
(add, sub)
}
|
