114 lines
3.1 KiB
Rust
114 lines
3.1 KiB
Rust
// Copyright 2015 The Rust Project Developers. See the COPYRIGHT
|
|
// file at the top-level directory of this distribution and at
|
|
// http://rust-lang.org/COPYRIGHT.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
|
|
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
|
|
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
|
|
// option. This file may not be copied, modified, or distributed
|
|
// except according to those terms.
|
|
|
|
#![feature(repr_simd, platform_intrinsics, concat_idents)]
|
|
#![allow(non_camel_case_types)]
|
|
|
|
use std::f32::NAN;
|
|
|
|
#[repr(simd)]
|
|
#[derive(Copy, Clone)]
|
|
struct i32x4(i32, i32, i32, i32);
|
|
#[repr(simd)]
|
|
#[derive(Copy, Clone)]
|
|
struct u32x4(pub u32, pub u32, pub u32, pub u32);
|
|
#[repr(simd)]
|
|
#[derive(Copy, Clone)]
|
|
struct f32x4(pub f32, pub f32, pub f32, pub f32);
|
|
|
|
extern "platform-intrinsic" {
|
|
fn simd_eq<T, U>(x: T, y: T) -> U;
|
|
fn simd_ne<T, U>(x: T, y: T) -> U;
|
|
fn simd_lt<T, U>(x: T, y: T) -> U;
|
|
fn simd_le<T, U>(x: T, y: T) -> U;
|
|
fn simd_gt<T, U>(x: T, y: T) -> U;
|
|
fn simd_ge<T, U>(x: T, y: T) -> U;
|
|
}
|
|
|
|
macro_rules! cmp {
|
|
($method: ident($lhs: expr, $rhs: expr)) => {{
|
|
let lhs = $lhs;
|
|
let rhs = $rhs;
|
|
let e: u32x4 = concat_idents!(simd_, $method)($lhs, $rhs);
|
|
// assume the scalar version is correct/the behaviour we want.
|
|
assert!((e.0 != 0) == lhs.0 .$method(&rhs.0));
|
|
assert!((e.1 != 0) == lhs.1 .$method(&rhs.1));
|
|
assert!((e.2 != 0) == lhs.2 .$method(&rhs.2));
|
|
assert!((e.3 != 0) == lhs.3 .$method(&rhs.3));
|
|
}}
|
|
}
|
|
macro_rules! tests {
|
|
($($lhs: ident, $rhs: ident;)*) => {{
|
|
$(
|
|
(|| {
|
|
cmp!(eq($lhs, $rhs));
|
|
cmp!(ne($lhs, $rhs));
|
|
|
|
// test both directions
|
|
cmp!(lt($lhs, $rhs));
|
|
cmp!(lt($rhs, $lhs));
|
|
|
|
cmp!(le($lhs, $rhs));
|
|
cmp!(le($rhs, $lhs));
|
|
|
|
cmp!(gt($lhs, $rhs));
|
|
cmp!(gt($rhs, $lhs));
|
|
|
|
cmp!(ge($lhs, $rhs));
|
|
cmp!(ge($rhs, $lhs));
|
|
})();
|
|
)*
|
|
}}
|
|
}
|
|
fn main() {
|
|
// 13 vs. -100 tests that we get signed vs. unsigned comparisons
|
|
// correct (i32: 13 > -100, u32: 13 < -100). let i1 = i32x4(10, -11, 12, 13);
|
|
let i1 = i32x4(10, -11, 12, 13);
|
|
let i2 = i32x4(5, -5, 20, -100);
|
|
let i3 = i32x4(10, -11, 20, -100);
|
|
|
|
let u1 = u32x4(10, !11+1, 12, 13);
|
|
let u2 = u32x4(5, !5+1, 20, !100+1);
|
|
let u3 = u32x4(10, !11+1, 20, !100+1);
|
|
|
|
let f1 = f32x4(10.0, -11.0, 12.0, 13.0);
|
|
let f2 = f32x4(5.0, -5.0, 20.0, -100.0);
|
|
let f3 = f32x4(10.0, -11.0, 20.0, -100.0);
|
|
|
|
unsafe {
|
|
tests! {
|
|
i1, i1;
|
|
u1, u1;
|
|
f1, f1;
|
|
|
|
i1, i2;
|
|
u1, u2;
|
|
f1, f2;
|
|
|
|
i1, i3;
|
|
u1, u3;
|
|
f1, f3;
|
|
}
|
|
}
|
|
|
|
// NAN comparisons are special:
|
|
// -11 (*) 13
|
|
// -5 -100 (*)
|
|
let f4 = f32x4(NAN, f1.1, NAN, f2.3);
|
|
|
|
unsafe {
|
|
tests! {
|
|
f1, f4;
|
|
f2, f4;
|
|
f4, f4;
|
|
}
|
|
}
|
|
}
|