Auto merge of rust-lang#138687 - calebzulawski:sync-from-portable-simd-2025-03-19, r=workingjubilee

Portable SIMD subtree update

Author: bors

library/portable-simd/beginners-guide.md

@@ -80,12 +80,12 @@ Most of the portable SIMD API is designed to allow the user to gloss over the de
 
 Fortunately, most SIMD types have a fairly predictable size. `i32x4` is bit-equivalent to `[i32; 4]` and so can be bitcast to it, e.g. using [`mem::transmute`], though the API usually offers a safe cast you can use instead.
 
-However, this is not the same as alignment. Computer architectures generally prefer aligned accesses, especially when moving data between memory and vector registers, and while some support specialized operations that can bend the rules to help with this, unaligned access is still typically slow, or even undefined behavior. In addition, different architectures can require different alignments when interacting with their native SIMD types. For this reason, any `#[repr(simd)]` type has a non-portable alignment. If it is necessary to directly interact with the alignment of these types, it should be via [`mem::align_of`].
+However, this is not the same as alignment. Computer architectures generally prefer aligned accesses, especially when moving data between memory and vector registers, and while some support specialized operations that can bend the rules to help with this, unaligned access is still typically slow, or even undefined behavior. In addition, different architectures can require different alignments when interacting with their native SIMD types. For this reason, any `#[repr(simd)]` type has a non-portable alignment. If it is necessary to directly interact with the alignment of these types, it should be via [`align_of`].
 
 When working with slices, data correctly aligned for SIMD can be acquired using the [`as_simd`] and [`as_simd_mut`] methods of the slice primitive.
 
 [`mem::transmute`]: https://doc.rust-lang.org/core/mem/fn.transmute.html
-[`mem::align_of`]: https://doc.rust-lang.org/core/mem/fn.align_of.html
+[`align_of`]: https://doc.rust-lang.org/core/mem/fn.align_of.html
 [`as_simd`]: https://doc.rust-lang.org/nightly/std/primitive.slice.html#method.as_simd
 [`as_simd_mut`]: https://doc.rust-lang.org/nightly/std/primitive.slice.html#method.as_simd_mut
 

library/portable-simd/crates/core_simd/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "core_simd"
 version = "0.1.0"
-edition = "2021"
+edition = "2024"
 homepage = "https://github.com/rust-lang/portable-simd"
 repository = "https://github.com/rust-lang/portable-simd"
 keywords = ["core", "simd", "intrinsics"]

library/portable-simd/crates/core_simd/src/lib.rs

@@ -35,7 +35,11 @@
     feature(stdarch_x86_avx512)
 )]
 #![warn(missing_docs, clippy::missing_inline_in_public_items)] // basically all items, really
-#![deny(unsafe_op_in_unsafe_fn, clippy::undocumented_unsafe_blocks)]
+#![deny(
+    unsafe_op_in_unsafe_fn,
+    unreachable_pub,
+    clippy::undocumented_unsafe_blocks
+)]
 #![doc(test(attr(deny(warnings))))]
 #![allow(internal_features)]
 #![unstable(feature = "portable_simd", issue = "86656")]

library/portable-simd/crates/core_simd/src/masks/bitmask.rs

@@ -5,7 +5,7 @@ use core::marker::PhantomData;
 
 /// A mask where each lane is represented by a single bit.
 #[repr(transparent)]
-pub struct Mask<T, const N: usize>(
+pub(crate) struct Mask<T, const N: usize>(
     <LaneCount<N> as SupportedLaneCount>::BitMask,
     PhantomData<T>,
 )
@@ -78,7 +78,7 @@ where
 {
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub fn splat(value: bool) -> Self {
+    pub(crate) fn splat(value: bool) -> Self {
         let mut mask = <LaneCount<N> as SupportedLaneCount>::BitMask::default();
         if value {
             mask.as_mut().fill(u8::MAX)
@@ -93,20 +93,20 @@ where
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub unsafe fn test_unchecked(&self, lane: usize) -> bool {
+    pub(crate) unsafe fn test_unchecked(&self, lane: usize) -> bool {
         (self.0.as_ref()[lane / 8] >> (lane % 8)) & 0x1 > 0
     }
 
     #[inline]
-    pub unsafe fn set_unchecked(&mut self, lane: usize, value: bool) {
+    pub(crate) unsafe fn set_unchecked(&mut self, lane: usize, value: bool) {
         unsafe {
             self.0.as_mut()[lane / 8] ^= ((value ^ self.test_unchecked(lane)) as u8) << (lane % 8)
         }
     }
 
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original value"]
-    pub fn to_int(self) -> Simd<T, N> {
+    pub(crate) fn to_int(self) -> Simd<T, N> {
         unsafe {
             core::intrinsics::simd::simd_select_bitmask(
                 self.0,
@@ -118,19 +118,19 @@ where
 
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub unsafe fn from_int_unchecked(value: Simd<T, N>) -> Self {
+    pub(crate) unsafe fn from_int_unchecked(value: Simd<T, N>) -> Self {
         unsafe { Self(core::intrinsics::simd::simd_bitmask(value), PhantomData) }
     }
 
     #[inline]
-    pub fn to_bitmask_integer(self) -> u64 {
+    pub(crate) fn to_bitmask_integer(self) -> u64 {
         let mut bitmask = [0u8; 8];
         bitmask[..self.0.as_ref().len()].copy_from_slice(self.0.as_ref());
         u64::from_ne_bytes(bitmask)
     }
 
     #[inline]
-    pub fn from_bitmask_integer(bitmask: u64) -> Self {
+    pub(crate) fn from_bitmask_integer(bitmask: u64) -> Self {
         let mut bytes = <LaneCount<N> as SupportedLaneCount>::BitMask::default();
         let len = bytes.as_mut().len();
         bytes
@@ -141,7 +141,7 @@ where
 
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub fn convert<U>(self) -> Mask<U, N>
+    pub(crate) fn convert<U>(self) -> Mask<U, N>
     where
         U: MaskElement,
     {
@@ -151,13 +151,13 @@ where
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub fn any(self) -> bool {
+    pub(crate) fn any(self) -> bool {
         self != Self::splat(false)
     }
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub fn all(self) -> bool {
+    pub(crate) fn all(self) -> bool {
         self == Self::splat(true)
     }
 }

library/portable-simd/crates/core_simd/src/masks/full_masks.rs

@@ -3,7 +3,7 @@
 use crate::simd::{LaneCount, MaskElement, Simd, SupportedLaneCount};
 
 #[repr(transparent)]
-pub struct Mask<T, const N: usize>(Simd<T, N>)
+pub(crate) struct Mask<T, const N: usize>(Simd<T, N>)
 where
     T: MaskElement,
     LaneCount<N>: SupportedLaneCount;
@@ -80,7 +80,7 @@ macro_rules! impl_reverse_bits {
             #[inline(always)]
             fn reverse_bits(self, n: usize) -> Self {
                 let rev = <$int>::reverse_bits(self);
-                let bitsize = core::mem::size_of::<$int>() * 8;
+                let bitsize = size_of::<$int>() * 8;
                 if n < bitsize {
                     // Shift things back to the right
                     rev >> (bitsize - n)
@@ -102,36 +102,36 @@ where
 {
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub fn splat(value: bool) -> Self {
+    pub(crate) fn splat(value: bool) -> Self {
         Self(Simd::splat(if value { T::TRUE } else { T::FALSE }))
     }
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub unsafe fn test_unchecked(&self, lane: usize) -> bool {
+    pub(crate) unsafe fn test_unchecked(&self, lane: usize) -> bool {
         T::eq(self.0[lane], T::TRUE)
     }
 
     #[inline]
-    pub unsafe fn set_unchecked(&mut self, lane: usize, value: bool) {
+    pub(crate) unsafe fn set_unchecked(&mut self, lane: usize, value: bool) {
         self.0[lane] = if value { T::TRUE } else { T::FALSE }
     }
 
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original value"]
-    pub fn to_int(self) -> Simd<T, N> {
+    pub(crate) fn to_int(self) -> Simd<T, N> {
         self.0
     }
 
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub unsafe fn from_int_unchecked(value: Simd<T, N>) -> Self {
+    pub(crate) unsafe fn from_int_unchecked(value: Simd<T, N>) -> Self {
         Self(value)
     }
 
     #[inline]
     #[must_use = "method returns a new mask and does not mutate the original value"]
-    pub fn convert<U>(self) -> Mask<U, N>
+    pub(crate) fn convert<U>(self) -> Mask<U, N>
     where
         U: MaskElement,
     {
@@ -220,14 +220,14 @@ where
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub fn any(self) -> bool {
+    pub(crate) fn any(self) -> bool {
         // Safety: use `self` as an integer vector
         unsafe { core::intrinsics::simd::simd_reduce_any(self.to_int()) }
     }
 
     #[inline]
     #[must_use = "method returns a new bool and does not mutate the original value"]
-    pub fn all(self) -> bool {
+    pub(crate) fn all(self) -> bool {
         // Safety: use `self` as an integer vector
         unsafe { core::intrinsics::simd::simd_reduce_all(self.to_int()) }
     }

library/portable-simd/crates/core_simd/src/ops.rs

@@ -1,4 +1,4 @@
-use crate::simd::{cmp::SimdPartialEq, LaneCount, Simd, SimdElement, SupportedLaneCount};
+use crate::simd::{LaneCount, Simd, SimdElement, SupportedLaneCount, cmp::SimdPartialEq};
 use core::ops::{Add, Mul};
 use core::ops::{BitAnd, BitOr, BitXor};
 use core::ops::{Div, Rem, Sub};

library/portable-simd/crates/core_simd/src/simd/cmp/eq.rs

@@ -1,6 +1,6 @@
 use crate::simd::{
-    ptr::{SimdConstPtr, SimdMutPtr},
     LaneCount, Mask, Simd, SimdElement, SupportedLaneCount,
+    ptr::{SimdConstPtr, SimdMutPtr},
 };
 
 /// Parallel `PartialEq`.

library/portable-simd/crates/core_simd/src/simd/cmp/ord.rs

@@ -1,7 +1,7 @@
 use crate::simd::{
+    LaneCount, Mask, Simd, SupportedLaneCount,
     cmp::SimdPartialEq,
     ptr::{SimdConstPtr, SimdMutPtr},
-    LaneCount, Mask, Simd, SupportedLaneCount,
 };
 
 /// Parallel `PartialOrd`.

library/portable-simd/crates/core_simd/src/simd/num/float.rs

@@ -1,7 +1,7 @@
 use super::sealed::Sealed;
 use crate::simd::{
-    cmp::{SimdPartialEq, SimdPartialOrd},
     LaneCount, Mask, Simd, SimdCast, SimdElement, SupportedLaneCount,
+    cmp::{SimdPartialEq, SimdPartialOrd},
 };
 
 /// Operations on SIMD vectors of floats.
@@ -263,7 +263,8 @@ macro_rules! impl_trait {
                 unsafe { core::intrinsics::simd::simd_as(self) }
             }
 
-            // https://github.com/llvm/llvm-project/issues/94694
+            // workaround for https://github.com/llvm/llvm-project/issues/94694 (fixed in LLVM 20)
+            // tracked in: https://github.com/rust-lang/rust/issues/135982
             #[cfg(target_arch = "aarch64")]
             #[inline]
             fn cast<T: SimdCast>(self) -> Self::Cast<T>
@@ -302,14 +303,14 @@ macro_rules! impl_trait {
 
             #[inline]
             fn to_bits(self) -> Simd<$bits_ty, N> {
-                assert_eq!(core::mem::size_of::<Self>(), core::mem::size_of::<Self::Bits>());
+                assert_eq!(size_of::<Self>(), size_of::<Self::Bits>());
                 // Safety: transmuting between vector types is safe
                 unsafe { core::mem::transmute_copy(&self) }
             }
 
             #[inline]
             fn from_bits(bits: Simd<$bits_ty, N>) -> Self {
-                assert_eq!(core::mem::size_of::<Self>(), core::mem::size_of::<Self::Bits>());
+                assert_eq!(size_of::<Self>(), size_of::<Self::Bits>());
                 // Safety: transmuting between vector types is safe
                 unsafe { core::mem::transmute_copy(&bits) }
             }

library/portable-simd/crates/core_simd/src/simd/num/int.rs

@@ -1,7 +1,7 @@
 use super::sealed::Sealed;
 use crate::simd::{
-    cmp::SimdOrd, cmp::SimdPartialOrd, num::SimdUint, LaneCount, Mask, Simd, SimdCast, SimdElement,
-    SupportedLaneCount,
+    LaneCount, Mask, Simd, SimdCast, SimdElement, SupportedLaneCount, cmp::SimdOrd,
+    cmp::SimdPartialOrd, num::SimdUint,
 };
 
 /// Operations on SIMD vectors of signed integers.

library/portable-simd/crates/core_simd/src/simd/num/uint.rs

@@ -1,5 +1,5 @@
 use super::sealed::Sealed;
-use crate::simd::{cmp::SimdOrd, LaneCount, Simd, SimdCast, SimdElement, SupportedLaneCount};
+use crate::simd::{LaneCount, Simd, SimdCast, SimdElement, SupportedLaneCount, cmp::SimdOrd};
 
 /// Operations on SIMD vectors of unsigned integers.
 pub trait SimdUint: Copy + Sealed {

library/portable-simd/crates/core_simd/src/simd/prelude.rs

@@ -7,10 +7,11 @@
 
 #[doc(no_inline)]
 pub use super::{
+    Mask, Simd,
     cmp::{SimdOrd, SimdPartialEq, SimdPartialOrd},
     num::{SimdFloat, SimdInt, SimdUint},
     ptr::{SimdConstPtr, SimdMutPtr},
-    simd_swizzle, Mask, Simd,
+    simd_swizzle,
 };
 
 #[rustfmt::skip]

library/portable-simd/crates/core_simd/src/simd/ptr/const_ptr.rs

@@ -1,5 +1,5 @@
 use super::sealed::Sealed;
-use crate::simd::{cmp::SimdPartialEq, num::SimdUint, LaneCount, Mask, Simd, SupportedLaneCount};
+use crate::simd::{LaneCount, Mask, Simd, SupportedLaneCount, cmp::SimdPartialEq, num::SimdUint};
 
 /// Operations on SIMD vectors of constant pointers.
 pub trait SimdConstPtr: Copy + Sealed {

library/portable-simd/crates/core_simd/src/simd/ptr/mut_ptr.rs

@@ -1,5 +1,5 @@
 use super::sealed::Sealed;
-use crate::simd::{cmp::SimdPartialEq, num::SimdUint, LaneCount, Mask, Simd, SupportedLaneCount};
+use crate::simd::{LaneCount, Mask, Simd, SupportedLaneCount, cmp::SimdPartialEq, num::SimdUint};
 
 /// Operations on SIMD vectors of mutable pointers.
 pub trait SimdMutPtr: Copy + Sealed {

library/portable-simd/crates/core_simd/src/swizzle.rs

@@ -214,6 +214,17 @@ where
     /// Rotates the vector such that the first `OFFSET` elements of the slice move to the end
     /// while the last `self.len() - OFFSET` elements move to the front. After calling `rotate_elements_left`,
     /// the element previously at index `OFFSET` will become the first element in the slice.
+    /// ```
+    /// # #![feature(portable_simd)]
+    /// # #[cfg(feature = "as_crate")] use core_simd::simd::Simd;
+    /// # #[cfg(not(feature = "as_crate"))] use core::simd::Simd;
+    /// let a = Simd::from_array([0, 1, 2, 3]);
+    /// let x = a.rotate_elements_left::<3>();
+    /// assert_eq!(x.to_array(), [3, 0, 1, 2]);
+    ///
+    /// let y = a.rotate_elements_left::<7>();
+    /// assert_eq!(y.to_array(), [3, 0, 1, 2]);
+    /// ```
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original inputs"]
     pub fn rotate_elements_left<const OFFSET: usize>(self) -> Self {
@@ -238,6 +249,17 @@ where
     /// Rotates the vector such that the first `self.len() - OFFSET` elements of the vector move to
     /// the end while the last `OFFSET` elements move to the front. After calling `rotate_elements_right`,
     /// the element previously at index `self.len() - OFFSET` will become the first element in the slice.
+    /// ```
+    /// # #![feature(portable_simd)]
+    /// # #[cfg(feature = "as_crate")] use core_simd::simd::Simd;
+    /// # #[cfg(not(feature = "as_crate"))] use core::simd::Simd;
+    /// let a = Simd::from_array([0, 1, 2, 3]);
+    /// let x = a.rotate_elements_right::<3>();
+    /// assert_eq!(x.to_array(), [1, 2, 3, 0]);
+    ///
+    /// let y = a.rotate_elements_right::<7>();
+    /// assert_eq!(y.to_array(), [1, 2, 3, 0]);
+    /// ```
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original inputs"]
     pub fn rotate_elements_right<const OFFSET: usize>(self) -> Self {
@@ -261,6 +283,17 @@ where
 
     /// Shifts the vector elements to the left by `OFFSET`, filling in with
     /// `padding` from the right.
+    /// ```
+    /// # #![feature(portable_simd)]
+    /// # #[cfg(feature = "as_crate")] use core_simd::simd::Simd;
+    /// # #[cfg(not(feature = "as_crate"))] use core::simd::Simd;
+    /// let a = Simd::from_array([0, 1, 2, 3]);
+    /// let x = a.shift_elements_left::<3>(255);
+    /// assert_eq!(x.to_array(), [3, 255, 255, 255]);
+    ///
+    /// let y = a.shift_elements_left::<7>(255);
+    /// assert_eq!(y.to_array(), [255, 255, 255, 255]);
+    /// ```
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original inputs"]
     pub fn shift_elements_left<const OFFSET: usize>(self, padding: T) -> Self {
@@ -283,6 +316,17 @@ where
 
     /// Shifts the vector elements to the right by `OFFSET`, filling in with
     /// `padding` from the left.
+    /// ```
+    /// # #![feature(portable_simd)]
+    /// # #[cfg(feature = "as_crate")] use core_simd::simd::Simd;
+    /// # #[cfg(not(feature = "as_crate"))] use core::simd::Simd;
+    /// let a = Simd::from_array([0, 1, 2, 3]);
+    /// let x = a.shift_elements_right::<3>(255);
+    /// assert_eq!(x.to_array(), [255, 255, 255, 0]);
+    ///
+    /// let y = a.shift_elements_right::<7>(255);
+    /// assert_eq!(y.to_array(), [255, 255, 255, 255]);
+    /// ```
     #[inline]
     #[must_use = "method returns a new vector and does not mutate the original inputs"]
     pub fn shift_elements_right<const OFFSET: usize>(self, padding: T) -> Self {

library/portable-simd/crates/core_simd/src/to_bytes.rs

@@ -1,6 +1,6 @@
 use crate::simd::{
-    num::{SimdFloat, SimdInt, SimdUint},
     LaneCount, Simd, SimdElement, SupportedLaneCount,
+    num::{SimdFloat, SimdInt, SimdUint},
 };
 
 mod sealed {