build(rust): vendor rnix parser crate

third_party/crates/hashbrown-0.14.5/src/raw/alloc.rs

@@ -0,0 +1,86 @@
+pub(crate) use self::inner::{do_alloc, Allocator, Global};
+
+// Nightly-case.
+// Use unstable `allocator_api` feature.
+// This is compatible with `allocator-api2` which can be enabled or not.
+// This is used when building for `std`.
+#[cfg(feature = "nightly")]
+mod inner {
+    use crate::alloc::alloc::Layout;
+    pub use crate::alloc::alloc::{Allocator, Global};
+    use core::ptr::NonNull;
+
+    #[allow(clippy::map_err_ignore)]
+    pub(crate) fn do_alloc<A: Allocator>(alloc: &A, layout: Layout) -> Result<NonNull<u8>, ()> {
+        match alloc.allocate(layout) {
+            Ok(ptr) => Ok(ptr.as_non_null_ptr()),
+            Err(_) => Err(()),
+        }
+    }
+}
+
+// Basic non-nightly case.
+// This uses `allocator-api2` enabled by default.
+// If any crate enables "nightly" in `allocator-api2`,
+// this will be equivalent to the nightly case,
+// since `allocator_api2::alloc::Allocator` would be re-export of
+// `core::alloc::Allocator`.
+#[cfg(all(not(feature = "nightly"), feature = "allocator-api2"))]
+mod inner {
+    use crate::alloc::alloc::Layout;
+    pub use allocator_api2::alloc::{Allocator, Global};
+    use core::ptr::NonNull;
+
+    #[allow(clippy::map_err_ignore)]
+    pub(crate) fn do_alloc<A: Allocator>(alloc: &A, layout: Layout) -> Result<NonNull<u8>, ()> {
+        match alloc.allocate(layout) {
+            Ok(ptr) => Ok(ptr.cast()),
+            Err(_) => Err(()),
+        }
+    }
+}
+
+// No-defaults case.
+// When building with default-features turned off and
+// neither `nightly` nor `allocator-api2` is enabled,
+// this will be used.
+// Making it impossible to use any custom allocator with collections defined
+// in this crate.
+// Any crate in build-tree can enable `allocator-api2`,
+// or `nightly` without disturbing users that don't want to use it.
+#[cfg(not(any(feature = "nightly", feature = "allocator-api2")))]
+mod inner {
+    use crate::alloc::alloc::{alloc, dealloc, Layout};
+    use core::ptr::NonNull;
+
+    #[allow(clippy::missing_safety_doc)] // not exposed outside of this crate
+    pub unsafe trait Allocator {
+        fn allocate(&self, layout: Layout) -> Result<NonNull<u8>, ()>;
+        unsafe fn deallocate(&self, ptr: NonNull<u8>, layout: Layout);
+    }
+
+    #[derive(Copy, Clone)]
+    pub struct Global;
+
+    unsafe impl Allocator for Global {
+        #[inline]
+        fn allocate(&self, layout: Layout) -> Result<NonNull<u8>, ()> {
+            unsafe { NonNull::new(alloc(layout)).ok_or(()) }
+        }
+        #[inline]
+        unsafe fn deallocate(&self, ptr: NonNull<u8>, layout: Layout) {
+            dealloc(ptr.as_ptr(), layout);
+        }
+    }
+
+    impl Default for Global {
+        #[inline]
+        fn default() -> Self {
+            Global
+        }
+    }
+
+    pub(crate) fn do_alloc<A: Allocator>(alloc: &A, layout: Layout) -> Result<NonNull<u8>, ()> {
+        alloc.allocate(layout)
+    }
+}

third_party/crates/hashbrown-0.14.5/src/raw/bitmask.rs

@@ -0,0 +1,133 @@
+use super::imp::{
+    BitMaskWord, NonZeroBitMaskWord, BITMASK_ITER_MASK, BITMASK_MASK, BITMASK_STRIDE,
+};
+
+/// A bit mask which contains the result of a `Match` operation on a `Group` and
+/// allows iterating through them.
+///
+/// The bit mask is arranged so that low-order bits represent lower memory
+/// addresses for group match results.
+///
+/// For implementation reasons, the bits in the set may be sparsely packed with
+/// groups of 8 bits representing one element. If any of these bits are non-zero
+/// then this element is considered to true in the mask. If this is the
+/// case, `BITMASK_STRIDE` will be 8 to indicate a divide-by-8 should be
+/// performed on counts/indices to normalize this difference. `BITMASK_MASK` is
+/// similarly a mask of all the actually-used bits.
+///
+/// To iterate over a bit mask, it must be converted to a form where only 1 bit
+/// is set per element. This is done by applying `BITMASK_ITER_MASK` on the
+/// mask bits.
+#[derive(Copy, Clone)]
+pub(crate) struct BitMask(pub(crate) BitMaskWord);
+
+#[allow(clippy::use_self)]
+impl BitMask {
+    /// Returns a new `BitMask` with all bits inverted.
+    #[inline]
+    #[must_use]
+    #[allow(dead_code)]
+    pub(crate) fn invert(self) -> Self {
+        BitMask(self.0 ^ BITMASK_MASK)
+    }
+
+    /// Returns a new `BitMask` with the lowest bit removed.
+    #[inline]
+    #[must_use]
+    fn remove_lowest_bit(self) -> Self {
+        BitMask(self.0 & (self.0 - 1))
+    }
+
+    /// Returns whether the `BitMask` has at least one set bit.
+    #[inline]
+    pub(crate) fn any_bit_set(self) -> bool {
+        self.0 != 0
+    }
+
+    /// Returns the first set bit in the `BitMask`, if there is one.
+    #[inline]
+    pub(crate) fn lowest_set_bit(self) -> Option<usize> {
+        if let Some(nonzero) = NonZeroBitMaskWord::new(self.0) {
+            Some(Self::nonzero_trailing_zeros(nonzero))
+        } else {
+            None
+        }
+    }
+
+    /// Returns the number of trailing zeroes in the `BitMask`.
+    #[inline]
+    pub(crate) fn trailing_zeros(self) -> usize {
+        // ARM doesn't have a trailing_zeroes instruction, and instead uses
+        // reverse_bits (RBIT) + leading_zeroes (CLZ). However older ARM
+        // versions (pre-ARMv7) don't have RBIT and need to emulate it
+        // instead. Since we only have 1 bit set in each byte on ARM, we can
+        // use swap_bytes (REV) + leading_zeroes instead.
+        if cfg!(target_arch = "arm") && BITMASK_STRIDE % 8 == 0 {
+            self.0.swap_bytes().leading_zeros() as usize / BITMASK_STRIDE
+        } else {
+            self.0.trailing_zeros() as usize / BITMASK_STRIDE
+        }
+    }
+
+    /// Same as above but takes a `NonZeroBitMaskWord`.
+    #[inline]
+    fn nonzero_trailing_zeros(nonzero: NonZeroBitMaskWord) -> usize {
+        if cfg!(target_arch = "arm") && BITMASK_STRIDE % 8 == 0 {
+            // SAFETY: A byte-swapped non-zero value is still non-zero.
+            let swapped = unsafe { NonZeroBitMaskWord::new_unchecked(nonzero.get().swap_bytes()) };
+            swapped.leading_zeros() as usize / BITMASK_STRIDE
+        } else {
+            nonzero.trailing_zeros() as usize / BITMASK_STRIDE
+        }
+    }
+
+    /// Returns the number of leading zeroes in the `BitMask`.
+    #[inline]
+    pub(crate) fn leading_zeros(self) -> usize {
+        self.0.leading_zeros() as usize / BITMASK_STRIDE
+    }
+}
+
+impl IntoIterator for BitMask {
+    type Item = usize;
+    type IntoIter = BitMaskIter;
+
+    #[inline]
+    fn into_iter(self) -> BitMaskIter {
+        // A BitMask only requires each element (group of bits) to be non-zero.
+        // However for iteration we need each element to only contain 1 bit.
+        BitMaskIter(BitMask(self.0 & BITMASK_ITER_MASK))
+    }
+}
+
+/// Iterator over the contents of a `BitMask`, returning the indices of set
+/// bits.
+#[derive(Copy, Clone)]
+pub(crate) struct BitMaskIter(pub(crate) BitMask);
+
+impl BitMaskIter {
+    /// Flip the bit in the mask for the entry at the given index.
+    ///
+    /// Returns the bit's previous state.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    #[cfg(feature = "raw")]
+    pub(crate) unsafe fn flip(&mut self, index: usize) -> bool {
+        // NOTE: The + BITMASK_STRIDE - 1 is to set the high bit.
+        let mask = 1 << (index * BITMASK_STRIDE + BITMASK_STRIDE - 1);
+        self.0 .0 ^= mask;
+        // The bit was set if the bit is now 0.
+        self.0 .0 & mask == 0
+    }
+}
+
+impl Iterator for BitMaskIter {
+    type Item = usize;
+
+    #[inline]
+    fn next(&mut self) -> Option<usize> {
+        let bit = self.0.lowest_set_bit()?;
+        self.0 = self.0.remove_lowest_bit();
+        Some(bit)
+    }
+}

third_party/crates/hashbrown-0.14.5/src/raw/generic.rs

@@ -0,0 +1,157 @@
+use super::bitmask::BitMask;
+use super::EMPTY;
+use core::{mem, ptr};
+
+// Use the native word size as the group size. Using a 64-bit group size on
+// a 32-bit architecture will just end up being more expensive because
+// shifts and multiplies will need to be emulated.
+
+cfg_if! {
+    if #[cfg(any(
+        target_pointer_width = "64",
+        target_arch = "aarch64",
+        target_arch = "x86_64",
+        target_arch = "wasm32",
+    ))] {
+        type GroupWord = u64;
+        type NonZeroGroupWord = core::num::NonZeroU64;
+    } else {
+        type GroupWord = u32;
+        type NonZeroGroupWord = core::num::NonZeroU32;
+    }
+}
+
+pub(crate) type BitMaskWord = GroupWord;
+pub(crate) type NonZeroBitMaskWord = NonZeroGroupWord;
+pub(crate) const BITMASK_STRIDE: usize = 8;
+// We only care about the highest bit of each byte for the mask.
+#[allow(clippy::cast_possible_truncation, clippy::unnecessary_cast)]
+pub(crate) const BITMASK_MASK: BitMaskWord = 0x8080_8080_8080_8080_u64 as GroupWord;
+pub(crate) const BITMASK_ITER_MASK: BitMaskWord = !0;
+
+/// Helper function to replicate a byte across a `GroupWord`.
+#[inline]
+fn repeat(byte: u8) -> GroupWord {
+    GroupWord::from_ne_bytes([byte; Group::WIDTH])
+}
+
+/// Abstraction over a group of control bytes which can be scanned in
+/// parallel.
+///
+/// This implementation uses a word-sized integer.
+#[derive(Copy, Clone)]
+pub(crate) struct Group(GroupWord);
+
+// We perform all operations in the native endianness, and convert to
+// little-endian just before creating a BitMask. The can potentially
+// enable the compiler to eliminate unnecessary byte swaps if we are
+// only checking whether a BitMask is empty.
+#[allow(clippy::use_self)]
+impl Group {
+    /// Number of bytes in the group.
+    pub(crate) const WIDTH: usize = mem::size_of::<Self>();
+
+    /// Returns a full group of empty bytes, suitable for use as the initial
+    /// value for an empty hash table.
+    ///
+    /// This is guaranteed to be aligned to the group size.
+    #[inline]
+    pub(crate) const fn static_empty() -> &'static [u8; Group::WIDTH] {
+        #[repr(C)]
+        struct AlignedBytes {
+            _align: [Group; 0],
+            bytes: [u8; Group::WIDTH],
+        }
+        const ALIGNED_BYTES: AlignedBytes = AlignedBytes {
+            _align: [],
+            bytes: [EMPTY; Group::WIDTH],
+        };
+        &ALIGNED_BYTES.bytes
+    }
+
+    /// Loads a group of bytes starting at the given address.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)] // unaligned load
+    pub(crate) unsafe fn load(ptr: *const u8) -> Self {
+        Group(ptr::read_unaligned(ptr.cast()))
+    }
+
+    /// Loads a group of bytes starting at the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn load_aligned(ptr: *const u8) -> Self {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        Group(ptr::read(ptr.cast()))
+    }
+
+    /// Stores the group of bytes to the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn store_aligned(self, ptr: *mut u8) {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        ptr::write(ptr.cast(), self.0);
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which *may*
+    /// have the given value.
+    ///
+    /// This function may return a false positive in certain cases where
+    /// the byte in the group differs from the searched value only in its
+    /// lowest bit. This is fine because:
+    /// - This never happens for `EMPTY` and `DELETED`, only full entries.
+    /// - The check for key equality will catch these.
+    /// - This only happens if there is at least 1 true match.
+    /// - The chance of this happening is very low (< 1% chance per byte).
+    #[inline]
+    pub(crate) fn match_byte(self, byte: u8) -> BitMask {
+        // This algorithm is derived from
+        // https://graphics.stanford.edu/~seander/bithacks.html##ValueInWord
+        let cmp = self.0 ^ repeat(byte);
+        BitMask((cmp.wrapping_sub(repeat(0x01)) & !cmp & repeat(0x80)).to_le())
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY`.
+    #[inline]
+    pub(crate) fn match_empty(self) -> BitMask {
+        // If the high bit is set, then the byte must be either:
+        // 1111_1111 (EMPTY) or 1000_0000 (DELETED).
+        // So we can just check if the top two bits are 1 by ANDing them.
+        BitMask((self.0 & (self.0 << 1) & repeat(0x80)).to_le())
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY` or `DELETED`.
+    #[inline]
+    pub(crate) fn match_empty_or_deleted(self) -> BitMask {
+        // A byte is EMPTY or DELETED iff the high bit is set
+        BitMask((self.0 & repeat(0x80)).to_le())
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are full.
+    #[inline]
+    pub(crate) fn match_full(self) -> BitMask {
+        self.match_empty_or_deleted().invert()
+    }
+
+    /// Performs the following transformation on all bytes in the group:
+    /// - `EMPTY => EMPTY`
+    /// - `DELETED => EMPTY`
+    /// - `FULL => DELETED`
+    #[inline]
+    pub(crate) fn convert_special_to_empty_and_full_to_deleted(self) -> Self {
+        // Map high_bit = 1 (EMPTY or DELETED) to 1111_1111
+        // and high_bit = 0 (FULL) to 1000_0000
+        //
+        // Here's this logic expanded to concrete values:
+        //   let full = 1000_0000 (true) or 0000_0000 (false)
+        //   !1000_0000 + 1 = 0111_1111 + 1 = 1000_0000 (no carry)
+        //   !0000_0000 + 0 = 1111_1111 + 0 = 1111_1111 (no carry)
+        let full = !self.0 & repeat(0x80);
+        Group(!full + (full >> 7))
+    }
+}

third_party/crates/hashbrown-0.14.5/src/raw/neon.rs

@@ -0,0 +1,124 @@
+use super::bitmask::BitMask;
+use super::EMPTY;
+use core::arch::aarch64 as neon;
+use core::mem;
+use core::num::NonZeroU64;
+
+pub(crate) type BitMaskWord = u64;
+pub(crate) type NonZeroBitMaskWord = NonZeroU64;
+pub(crate) const BITMASK_STRIDE: usize = 8;
+pub(crate) const BITMASK_MASK: BitMaskWord = !0;
+pub(crate) const BITMASK_ITER_MASK: BitMaskWord = 0x8080_8080_8080_8080;
+
+/// Abstraction over a group of control bytes which can be scanned in
+/// parallel.
+///
+/// This implementation uses a 64-bit NEON value.
+#[derive(Copy, Clone)]
+pub(crate) struct Group(neon::uint8x8_t);
+
+#[allow(clippy::use_self)]
+impl Group {
+    /// Number of bytes in the group.
+    pub(crate) const WIDTH: usize = mem::size_of::<Self>();
+
+    /// Returns a full group of empty bytes, suitable for use as the initial
+    /// value for an empty hash table.
+    ///
+    /// This is guaranteed to be aligned to the group size.
+    #[inline]
+    pub(crate) const fn static_empty() -> &'static [u8; Group::WIDTH] {
+        #[repr(C)]
+        struct AlignedBytes {
+            _align: [Group; 0],
+            bytes: [u8; Group::WIDTH],
+        }
+        const ALIGNED_BYTES: AlignedBytes = AlignedBytes {
+            _align: [],
+            bytes: [EMPTY; Group::WIDTH],
+        };
+        &ALIGNED_BYTES.bytes
+    }
+
+    /// Loads a group of bytes starting at the given address.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)] // unaligned load
+    pub(crate) unsafe fn load(ptr: *const u8) -> Self {
+        Group(neon::vld1_u8(ptr))
+    }
+
+    /// Loads a group of bytes starting at the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn load_aligned(ptr: *const u8) -> Self {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        Group(neon::vld1_u8(ptr))
+    }
+
+    /// Stores the group of bytes to the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn store_aligned(self, ptr: *mut u8) {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        neon::vst1_u8(ptr, self.0);
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which *may*
+    /// have the given value.
+    #[inline]
+    pub(crate) fn match_byte(self, byte: u8) -> BitMask {
+        unsafe {
+            let cmp = neon::vceq_u8(self.0, neon::vdup_n_u8(byte));
+            BitMask(neon::vget_lane_u64(neon::vreinterpret_u64_u8(cmp), 0))
+        }
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY`.
+    #[inline]
+    pub(crate) fn match_empty(self) -> BitMask {
+        self.match_byte(EMPTY)
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY` or `DELETED`.
+    #[inline]
+    pub(crate) fn match_empty_or_deleted(self) -> BitMask {
+        unsafe {
+            let cmp = neon::vcltz_s8(neon::vreinterpret_s8_u8(self.0));
+            BitMask(neon::vget_lane_u64(neon::vreinterpret_u64_u8(cmp), 0))
+        }
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are full.
+    #[inline]
+    pub(crate) fn match_full(self) -> BitMask {
+        unsafe {
+            let cmp = neon::vcgez_s8(neon::vreinterpret_s8_u8(self.0));
+            BitMask(neon::vget_lane_u64(neon::vreinterpret_u64_u8(cmp), 0))
+        }
+    }
+
+    /// Performs the following transformation on all bytes in the group:
+    /// - `EMPTY => EMPTY`
+    /// - `DELETED => EMPTY`
+    /// - `FULL => DELETED`
+    #[inline]
+    pub(crate) fn convert_special_to_empty_and_full_to_deleted(self) -> Self {
+        // Map high_bit = 1 (EMPTY or DELETED) to 1111_1111
+        // and high_bit = 0 (FULL) to 1000_0000
+        //
+        // Here's this logic expanded to concrete values:
+        //   let special = 0 > byte = 1111_1111 (true) or 0000_0000 (false)
+        //   1111_1111 | 1000_0000 = 1111_1111
+        //   0000_0000 | 1000_0000 = 1000_0000
+        unsafe {
+            let special = neon::vcltz_s8(neon::vreinterpret_s8_u8(self.0));
+            Group(neon::vorr_u8(special, neon::vdup_n_u8(0x80)))
+        }
+    }
+}

third_party/crates/hashbrown-0.14.5/src/raw/sse2.rs

@@ -0,0 +1,149 @@
+use super::bitmask::BitMask;
+use super::EMPTY;
+use core::mem;
+use core::num::NonZeroU16;
+
+#[cfg(target_arch = "x86")]
+use core::arch::x86;
+#[cfg(target_arch = "x86_64")]
+use core::arch::x86_64 as x86;
+
+pub(crate) type BitMaskWord = u16;
+pub(crate) type NonZeroBitMaskWord = NonZeroU16;
+pub(crate) const BITMASK_STRIDE: usize = 1;
+pub(crate) const BITMASK_MASK: BitMaskWord = 0xffff;
+pub(crate) const BITMASK_ITER_MASK: BitMaskWord = !0;
+
+/// Abstraction over a group of control bytes which can be scanned in
+/// parallel.
+///
+/// This implementation uses a 128-bit SSE value.
+#[derive(Copy, Clone)]
+pub(crate) struct Group(x86::__m128i);
+
+// FIXME: https://github.com/rust-lang/rust-clippy/issues/3859
+#[allow(clippy::use_self)]
+impl Group {
+    /// Number of bytes in the group.
+    pub(crate) const WIDTH: usize = mem::size_of::<Self>();
+
+    /// Returns a full group of empty bytes, suitable for use as the initial
+    /// value for an empty hash table.
+    ///
+    /// This is guaranteed to be aligned to the group size.
+    #[inline]
+    #[allow(clippy::items_after_statements)]
+    pub(crate) const fn static_empty() -> &'static [u8; Group::WIDTH] {
+        #[repr(C)]
+        struct AlignedBytes {
+            _align: [Group; 0],
+            bytes: [u8; Group::WIDTH],
+        }
+        const ALIGNED_BYTES: AlignedBytes = AlignedBytes {
+            _align: [],
+            bytes: [EMPTY; Group::WIDTH],
+        };
+        &ALIGNED_BYTES.bytes
+    }
+
+    /// Loads a group of bytes starting at the given address.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)] // unaligned load
+    pub(crate) unsafe fn load(ptr: *const u8) -> Self {
+        Group(x86::_mm_loadu_si128(ptr.cast()))
+    }
+
+    /// Loads a group of bytes starting at the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn load_aligned(ptr: *const u8) -> Self {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        Group(x86::_mm_load_si128(ptr.cast()))
+    }
+
+    /// Stores the group of bytes to the given address, which must be
+    /// aligned to `mem::align_of::<Group>()`.
+    #[inline]
+    #[allow(clippy::cast_ptr_alignment)]
+    pub(crate) unsafe fn store_aligned(self, ptr: *mut u8) {
+        // FIXME: use align_offset once it stabilizes
+        debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0);
+        x86::_mm_store_si128(ptr.cast(), self.0);
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which have
+    /// the given value.
+    #[inline]
+    pub(crate) fn match_byte(self, byte: u8) -> BitMask {
+        #[allow(
+            clippy::cast_possible_wrap, // byte: u8 as i8
+            // byte: i32 as u16
+            //   note: _mm_movemask_epi8 returns a 16-bit mask in a i32, the
+            //   upper 16-bits of the i32 are zeroed:
+            clippy::cast_sign_loss,
+            clippy::cast_possible_truncation
+        )]
+        unsafe {
+            let cmp = x86::_mm_cmpeq_epi8(self.0, x86::_mm_set1_epi8(byte as i8));
+            BitMask(x86::_mm_movemask_epi8(cmp) as u16)
+        }
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY`.
+    #[inline]
+    pub(crate) fn match_empty(self) -> BitMask {
+        self.match_byte(EMPTY)
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are
+    /// `EMPTY` or `DELETED`.
+    #[inline]
+    pub(crate) fn match_empty_or_deleted(self) -> BitMask {
+        #[allow(
+            // byte: i32 as u16
+            //   note: _mm_movemask_epi8 returns a 16-bit mask in a i32, the
+            //   upper 16-bits of the i32 are zeroed:
+            clippy::cast_sign_loss,
+            clippy::cast_possible_truncation
+        )]
+        unsafe {
+            // A byte is EMPTY or DELETED iff the high bit is set
+            BitMask(x86::_mm_movemask_epi8(self.0) as u16)
+        }
+    }
+
+    /// Returns a `BitMask` indicating all bytes in the group which are full.
+    #[inline]
+    pub(crate) fn match_full(&self) -> BitMask {
+        self.match_empty_or_deleted().invert()
+    }
+
+    /// Performs the following transformation on all bytes in the group:
+    /// - `EMPTY => EMPTY`
+    /// - `DELETED => EMPTY`
+    /// - `FULL => DELETED`
+    #[inline]
+    pub(crate) fn convert_special_to_empty_and_full_to_deleted(self) -> Self {
+        // Map high_bit = 1 (EMPTY or DELETED) to 1111_1111
+        // and high_bit = 0 (FULL) to 1000_0000
+        //
+        // Here's this logic expanded to concrete values:
+        //   let special = 0 > byte = 1111_1111 (true) or 0000_0000 (false)
+        //   1111_1111 | 1000_0000 = 1111_1111
+        //   0000_0000 | 1000_0000 = 1000_0000
+        #[allow(
+            clippy::cast_possible_wrap, // byte: 0x80_u8 as i8
+        )]
+        unsafe {
+            let zero = x86::_mm_setzero_si128();
+            let special = x86::_mm_cmpgt_epi8(zero, self.0);
+            Group(x86::_mm_or_si128(
+                special,
+                x86::_mm_set1_epi8(0x80_u8 as i8),
+            ))
+        }
+    }
+}