Skip to content

cpufeatures: add support for avx10.1 and avx10.2 #1531

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
MuntasirSZN wants to merge 1 commit into
base: master
Choose a base branch
from
Open

cpufeatures: add support for avx10.1 and avx10.2 #1531

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions cpufeatures/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -101,6 +101,8 @@ Target features:
- `adx`
- `aes`
- `avx`
- `avx10.1`*
- `avx10.2`*
- `avx2`
- `avx512bw`*
- `avx512cd`*
Expand Down
1 change: 1 addition & 0 deletions cpufeatures/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@ compile_error!("This crate works only on `aarch64`, `loongarch64`, `x86`, and `x
#[macro_export]
macro_rules! new {
($mod_name:ident, $($tf:tt),+ $(,)?) => {
#[allow(unexpected_cfgs)]
mod $mod_name {
use core::sync::atomic::{AtomicU8, Ordering::Relaxed};

Expand Down
25 changes: 24 additions & 1 deletion cpufeatures/src/x86.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,13 @@ macro_rules! __detect_target_features {
}

let cr = unsafe {
[cpuid(1), cpuid_count(7, 0), cpuid_count(7, 1)]
[
cpuid(1),
cpuid_count(7, 0),
cpuid_count(7, 1),
cpuid(0),
cpuid_count(0x24, 0),
]
};

$($crate::check!(cr, $tf) & )+ true
Expand Down Expand Up @@ -81,6 +87,23 @@ macro_rules! __expand_check_macro {
#[macro_export]
#[doc(hidden)]
macro_rules! check {
($cr:expr, "avx10.1") => {{
$crate::__xgetbv!($cr, 0b1110_0110)
& ($cr[3].eax >= 0x24)
& ($cr[2].edx & (1 << 19) != 0)
& (($cr[4].ebx & 0xff) >= 1)
& (($cr[4].ebx & (1 << 18)) != 0)
}};
($cr:expr, "avx10.2") => {{
$crate::__xgetbv!($cr, 0b1110_0110)
& ($cr[3].eax >= 0x24)
& ($cr[2].edx & (1 << 19) != 0)
& (($cr[4].ebx & 0xff) >= 2)
& (($cr[4].ebx & (1 << 18)) != 0)
& ($cr[2].eax & (1 << 4) != 0)
& ($cr[2].edx & (1 << 4) != 0)
& ($cr[2].edx & (1 << 10) != 0)
}};
$(
($cr:expr, $name) => {{
// Register bits are listed here:
Expand Down
168 changes: 168 additions & 0 deletions cpufeatures/tests/x86.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
#![cfg(any(target_arch = "x86", target_arch = "x86_64"))]

cpufeatures::new!(cpuid, "aes", "sha");
cpufeatures::new!(cpuid_avx10, "avx10.1", "avx10.2");

#[test]
fn init() {
Expand All @@ -15,3 +16,170 @@ fn init_get() {
let (token, val) = cpuid::init_get();
assert_eq!(val, token.get());
}

#[test]
fn avx10_probe_does_not_fault() {
let (token, val) = cpuid_avx10::init_get();
assert_eq!(val, token.get());
let _ = cpuid_avx10::get();
}

#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
mod avx10_logic {
#[cfg(target_arch = "x86")]
use core::arch::x86::CpuidResult;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64::CpuidResult;

fn leaf1() -> CpuidResult {
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0b11 << 26,
edx: 0,
}
}

fn blank() -> CpuidResult {
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0,
edx: 0,
}
}

#[test]
fn version_zero_is_disabled() {
let cr = [
leaf1(),
blank(),
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0,
edx: 1 << 19,
},
CpuidResult {
eax: 0x24,
ebx: 0,
ecx: 0,
edx: 0,
},
CpuidResult {
eax: 0,
ebx: (1 << 17) | (1 << 18),
ecx: 0,
edx: 0,
},
];
assert!(!cpufeatures::check!(cr, "avx10.1"));
assert!(!cpufeatures::check!(cr, "avx10.2"));
}

#[test]
fn max_leaf_guard_holds() {
let cr = [
leaf1(),
blank(),
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0,
edx: 1 << 19,
},
CpuidResult {
eax: 0x16,
ebx: 0,
ecx: 0,
edx: 0,
},
CpuidResult {
eax: 0,
ebx: 1 | (1 << 17) | (1 << 18),
ecx: 0,
edx: 0,
},
];
assert!(!cpufeatures::check!(cr, "avx10.1"));
assert!(!cpufeatures::check!(cr, "avx10.2"));
}

#[test]
fn presence_bit_required() {
let cr = [
leaf1(),
blank(),
blank(),
CpuidResult {
eax: 0x24,
ebx: 0,
ecx: 0,
edx: 0,
},
CpuidResult {
eax: 0,
ebx: 1 | (1 << 17) | (1 << 18),
ecx: 0,
edx: 0,
},
];
assert!(!cpufeatures::check!(cr, "avx10.1"));
assert!(!cpufeatures::check!(cr, "avx10.2"));
}

#[test]
fn narrow_vector_without_512_bit_reports_false() {
let cr = [
leaf1(),
blank(),
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0,
edx: 1 << 19,
},
CpuidResult {
eax: 0x24,
ebx: 0,
ecx: 0,
edx: 0,
},
CpuidResult {
eax: 0,
ebx: 1 | (1 << 17),
ecx: 0,
edx: 0,
},
];
assert!(!cpufeatures::check!(cr, "avx10.1"));
assert!(!cpufeatures::check!(cr, "avx10.2"));
}

#[test]
fn vnni_trio_required_for_avx10_2() {
let cr = [
leaf1(),
blank(),
CpuidResult {
eax: 0,
ebx: 0,
ecx: 0,
edx: 1 << 19,
},
CpuidResult {
eax: 0x24,
ebx: 0,
ecx: 0,
edx: 0,
},
CpuidResult {
eax: 0,
ebx: 2 | (1 << 18),
ecx: 0,
edx: 0,
},
];
assert!(!cpufeatures::check!(cr, "avx10.2"));
}
}