Skip to content

Ensure same behavior on all supported CPU archs #1174

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

Merged
brunocodutra merged 1 commit into from
Aug 16, 2026
Merged

Ensure same behavior on all supported CPU archs #1174

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: 1 addition & 1 deletion Cargo.lock
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
[package]
name = "cinder"
version = "0.6.0"
version = "0.6.1"
authors = ["Bruno Dutra <brunocodutra@gmail.com>"]
edition = "2024"
description = "A hobby chess engine"
Expand Down
2 changes: 1 addition & 1 deletion Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
CRATE_VERSION := 0.6.0
CRATE_VERSION := 0.6.1
TARGET_DIR := $(CURDIR)/target
BIN_DIR := $(TARGET_DIR)/bin

Expand Down
2 changes: 1 addition & 1 deletion README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,7 +93,7 @@ Cinder is distributed under the terms of the GPL-3.0 license, see [LICENSE] for
[Leela Chess Zero]: https://lczero.org/
[Open Database License]: https://opendatacommons.org/licenses/odbl/1-0/

[v0.6.*]: https://github.com/brunocodutra/cinder/releases/tag/v0.6.0
[v0.6.*]: https://github.com/brunocodutra/cinder/releases/tag/v0.6.1
[v0.5.*]: https://github.com/brunocodutra/cinder/releases/tag/v0.5.2
[v0.4.*]: https://github.com/brunocodutra/cinder/releases/tag/v0.4.1
[v0.3.*]: https://github.com/brunocodutra/cinder/releases/tag/v0.3.1
Expand Down
9 changes: 4 additions & 5 deletions lib/chess/pins.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,16 +35,15 @@ impl Pins {
};
}

let piece_mask = board.squares()[turn].to_simd().find(
let pins = furled.mask(pinned).extend().mask(beyond).unfurl(rays);
let pinned = board.squares()[turn].to_simd().find(
rays.compress(pinned.to_bitmask()).extract::<0, 16>(),
pinned.count().cast(),
);

let pinned = furled.mask(pinned).extend().mask(beyond).unfurl(rays);

Pins {
attacks: threats[turn] & (u16x64::splat(!piece_mask) | pinned.to_idx_set()),
unpinned: !pinned.occupied(),
attacks: threats[turn] & (u16x64::splat(!pinned) | pins.to_idx_set()),
unpinned: !pins.occupied(),
}
}

Expand Down
18 changes: 9 additions & 9 deletions lib/chess/placement.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -635,21 +635,21 @@ impl Placement {

#[cfg(target_feature = "avx512f")]
{
let idx_set = u16x64::splat(1).shlv(indices.cast::<u16>());
indices.simd_ne(zeroed()).select(idx_set, zeroed())
let ones = self.occupied().select(Simd::splat(1), Simd::splat(0));
ones.shlv(indices.cast::<u16>())
}

#[cfg(not(target_feature = "avx512f"))]
unsafe {
#[rustfmt::skip]
const S0: u8x64 = Simd::from_array([
0x00, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
]);

Expand All @@ -665,9 +665,9 @@ impl Placement {
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80,
]);

transmute_copy::<(u8x64, u8x64), u16x64>(
&S0.shuffle(indices).interleave(S1.shuffle(indices)),
)
let (left, right) = S0.shuffle(indices).interleave(S1.shuffle(indices));
let indices = transmute_copy::<[u8x64; 2], u16x64>(&[left, right]);
self.occupied().select(indices, zeroed())
}
}

Expand Down
140 changes: 10 additions & 130 deletions lib/simd/shlv.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,142 +1,19 @@
use crate::{simd::Halve, util::Num};
use std::simd::prelude::*;
use crate::util::{Assume, Num};
use std::{array, simd::prelude::*};

/// Trait for [`Simd<_, _>` ] types that can shift left dynamically.
pub trait Shlv {
/// Shifts left by `shift`.
fn shlv(self, shift: Self) -> Self;
}

impl Shlv for u16x64 {
impl<const N: usize> Shlv for Simd<u16, N> {
#[inline(always)]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

let [x0, x1] = self.halve();
let [s0, s1] = shift.halve();
Halve::merge([x0.shlv(s0), x1.shlv(s1)])
}
}

impl Shlv for u16x32 {
#[inline(always)]
#[cfg(target_feature = "avx512bw")]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
use std::arch::x86_64::*;
_mm512_sllv_epi16(self.into(), shift.into()).into()
}
}

#[inline(always)]
#[cfg(not(target_feature = "avx512bw"))]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
let [x0, x1] = self.halve();
let [s0, s1] = shift.halve();
Halve::merge([x0.shlv(s0), x1.shlv(s1)])
}
}

impl Shlv for u16x16 {
#[inline(always)]
#[cfg(all(target_feature = "avx512bw", target_feature = "avx512vl"))]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
use std::arch::x86_64::*;
_mm256_sllv_epi16(self.into(), shift.into()).into()
}
}

#[inline(always)]
#[cfg(not(all(target_feature = "avx512bw", target_feature = "avx512vl")))]
#[cfg(target_feature = "avx2")]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
use std::arch::x86_64::*;

let x0 = _mm256_unpacklo_epi16(Simd::splat(0).into(), self.into());
let x1 = _mm256_unpackhi_epi16(Simd::splat(0).into(), self.into());
let s0 = _mm256_unpacklo_epi16(shift.into(), Simd::splat(0).into());
let s1 = _mm256_unpackhi_epi16(shift.into(), Simd::splat(0).into());
let r0 = _mm256_srli_epi32(_mm256_sllv_epi32(x0, s0), 16);
let r1 = _mm256_srli_epi32(_mm256_sllv_epi32(x1, s1), 16);
_mm256_packus_epi32(r0, r1).into()
}
}

#[inline(always)]
#[cfg(not(all(target_feature = "avx512bw", target_feature = "avx512vl")))]
#[cfg(not(target_feature = "avx2"))]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
let [x0, x1] = self.halve();
let [s0, s1] = shift.halve();
Halve::merge([x0.shlv(s0), x1.shlv(s1)])
}
}

impl Shlv for u16x8 {
#[inline(always)]
#[cfg(all(target_feature = "avx512bw", target_feature = "avx512vl"))]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
use std::arch::x86_64::*;
_mm_sllv_epi16(self.into(), shift.into()).into()
}
}

#[inline(always)]
#[cfg(target_feature = "neon")]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
use std::arch::aarch64::*;
vshlq_u16(self.into(), vreinterpretq_s16_u16(shift.into())).into()
}
}

#[inline(always)]
#[cfg(not(all(target_feature = "avx512bw", target_feature = "avx512vl")))]
#[cfg(not(target_feature = "neon"))]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn shlv(self, shift: Self) -> Self {
fallback(self, shift)
}
}

#[allow(unused)]
#[inline(always)]
#[cfg_attr(feature = "no_panic", no_panic::no_panic)]
fn fallback(x: u16x8, shift: u16x8) -> u16x8 {
debug_assert!(shift.simd_lt(Simd::splat(16)).all());

unsafe {
u16x8::from_array([
x[0].unchecked_shl(shift[0].cast()),
x[1].unchecked_shl(shift[1].cast()),
x[2].unchecked_shl(shift[2].cast()),
x[3].unchecked_shl(shift[3].cast()),
x[4].unchecked_shl(shift[4].cast()),
x[5].unchecked_shl(shift[5].cast()),
x[6].unchecked_shl(shift[6].cast()),
x[7].unchecked_shl(shift[7].cast()),
])
Simd::from_array(array::from_fn(|i| {
self[i].checked_shl(shift[i].cast()).assume()
}))
}
}

Expand Down Expand Up @@ -191,6 +68,9 @@ mod tests {
#[strategy(UniformArrayStrategy::new(0u16..=255u16).prop_map(u16x8::from_array))] x: u16x8,
#[strategy(UniformArrayStrategy::new(0u16..8u16).prop_map(u16x8::from_array))] s: u16x8,
) {
assert_eq!(x.shlv(s), fallback(x, s));
use crate::simd::Halve;
let [x0, x1] = x.halve();
let [s0, s1] = s.halve();
assert_eq!(x.shlv(s).halve(), [x0.shlv(s0), x1.shlv(s1)]);
}
}
Loading