diff --git a/crates/core_simd/src/cast.rs b/crates/core_simd/src/cast.rs
new file mode 100644
index 00000000000..e04a9042b1b
--- /dev/null
+++ b/crates/core_simd/src/cast.rs
@@ -0,0 +1,45 @@
+use crate::simd::SimdElement;
+
+/// Supporting trait for `Simd::cast`.  Typically doesn't need to be used directly.
+pub trait SimdCast<Target: SimdElement>: SimdElement {}
+
+macro_rules! into_number {
+    { $($type:ty),* } => {
+        $(
+        impl SimdCast<i8> for $type {}
+        impl SimdCast<i16> for $type {}
+        impl SimdCast<i32> for $type {}
+        impl SimdCast<i64> for $type {}
+        impl SimdCast<isize> for $type {}
+
+        impl SimdCast<u8> for $type {}
+        impl SimdCast<u16> for $type {}
+        impl SimdCast<u32> for $type {}
+        impl SimdCast<u64> for $type {}
+        impl SimdCast<usize> for $type {}
+
+        impl SimdCast<f32> for $type {}
+        impl SimdCast<f64> for $type {}
+        )*
+    }
+}
+
+into_number! { i8, i16, i32, i64, isize, u8, u16, u32, u64, usize, f32, f64 }
+
+macro_rules! into_pointer {
+    { $($type:ty),* } => {
+        $(
+        impl<T> SimdCast<$type> for *const T {}
+        impl<T> SimdCast<$type> for *mut T {}
+        impl<T> SimdCast<*const T> for $type {}
+        impl<T> SimdCast<*mut T> for $type {}
+        )*
+    }
+}
+
+into_pointer! { i8, i16, i32, i64, isize, u8, u16, u32, u64, usize }
+
+impl<T, U> SimdCast<*const T> for *const U {}
+impl<T, U> SimdCast<*const T> for *mut U {}
+impl<T, U> SimdCast<*mut T> for *const U {}
+impl<T, U> SimdCast<*mut T> for *mut U {}
diff --git a/crates/core_simd/src/elements.rs b/crates/core_simd/src/elements.rs
index 701eb66b248..dc7f52a4d57 100644
--- a/crates/core_simd/src/elements.rs
+++ b/crates/core_simd/src/elements.rs
@@ -1,11 +1,15 @@
+mod const_ptr;
 mod float;
 mod int;
+mod mut_ptr;
 mod uint;
 
 mod sealed {
     pub trait Sealed {}
 }
 
+pub use const_ptr::*;
 pub use float::*;
 pub use int::*;
+pub use mut_ptr::*;
 pub use uint::*;
diff --git a/crates/core_simd/src/elements/const_ptr.rs b/crates/core_simd/src/elements/const_ptr.rs
new file mode 100644
index 00000000000..ab6b5b8b5f4
--- /dev/null
+++ b/crates/core_simd/src/elements/const_ptr.rs
@@ -0,0 +1,59 @@
+use super::sealed::Sealed;
+use crate::simd::{intrinsics, LaneCount, Mask, Simd, SimdPartialEq, SupportedLaneCount};
+
+/// Operations on SIMD vectors of constant pointers.
+pub trait SimdConstPtr: Copy + Sealed {
+    /// Vector type representing the pointers as bits.
+    type Bits;
+
+    /// Vector of mutable pointers to the same type.
+    type MutPtr;
+
+    /// Mask type used for manipulating this SIMD vector type.
+    type Mask;
+
+    /// Returns `true` for each lane that is null.
+    fn is_null(self) -> Self::Mask;
+
+    /// Changes constness without changing the type.
+    fn as_mut(self) -> Self::MutPtr;
+
+    /// Cast pointers to raw bits.
+    fn to_bits(self) -> Self::Bits;
+
+    /// Cast raw bits to pointers.
+    fn from_bits(bits: Self::Bits) -> Self;
+}
+
+impl<T, const LANES: usize> Sealed for Simd<*const T, LANES> where
+    LaneCount<LANES>: SupportedLaneCount
+{
+}
+
+impl<T, const LANES: usize> SimdConstPtr for Simd<*const T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    type Bits = Simd<usize, LANES>;
+    type MutPtr = Simd<*mut T, LANES>;
+    type Mask = Mask<isize, LANES>;
+
+    fn is_null(self) -> Self::Mask {
+        Simd::splat(core::ptr::null()).simd_eq(self)
+    }
+
+    fn as_mut(self) -> Self::MutPtr {
+        // Converting between pointers is safe
+        unsafe { intrinsics::simd_as(self) }
+    }
+
+    fn to_bits(self) -> Self::Bits {
+        // Casting pointers to usize is safe
+        unsafe { intrinsics::simd_as(self) }
+    }
+
+    fn from_bits(bits: Self::Bits) -> Self {
+        // Casting usize to pointers is safe
+        unsafe { intrinsics::simd_as(bits) }
+    }
+}
diff --git a/crates/core_simd/src/elements/mut_ptr.rs b/crates/core_simd/src/elements/mut_ptr.rs
new file mode 100644
index 00000000000..b49f9fda7e4
--- /dev/null
+++ b/crates/core_simd/src/elements/mut_ptr.rs
@@ -0,0 +1,57 @@
+use super::sealed::Sealed;
+use crate::simd::{intrinsics, LaneCount, Mask, Simd, SimdPartialEq, SupportedLaneCount};
+
+/// Operations on SIMD vectors of mutable pointers.
+pub trait SimdMutPtr: Copy + Sealed {
+    /// Vector type representing the pointers as bits.
+    type Bits;
+
+    /// Vector of constant pointers to the same type.
+    type ConstPtr;
+
+    /// Mask type used for manipulating this SIMD vector type.
+    type Mask;
+
+    /// Returns `true` for each lane that is null.
+    fn is_null(self) -> Self::Mask;
+
+    /// Changes constness without changing the type.
+    fn as_const(self) -> Self::ConstPtr;
+
+    /// Cast pointers to raw bits.
+    fn to_bits(self) -> Self::Bits;
+
+    /// Cast raw bits to pointers.
+    fn from_bits(bits: Self::Bits) -> Self;
+}
+
+impl<T, const LANES: usize> Sealed for Simd<*mut T, LANES> where LaneCount<LANES>: SupportedLaneCount
+{}
+
+impl<T, const LANES: usize> SimdMutPtr for Simd<*mut T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    type Bits = Simd<usize, LANES>;
+    type ConstPtr = Simd<*const T, LANES>;
+    type Mask = Mask<isize, LANES>;
+
+    fn is_null(self) -> Self::Mask {
+        Simd::splat(core::ptr::null_mut()).simd_eq(self)
+    }
+
+    fn as_const(self) -> Self::ConstPtr {
+        // Converting between pointers is safe
+        unsafe { intrinsics::simd_as(self) }
+    }
+
+    fn to_bits(self) -> Self::Bits {
+        // Casting pointers to usize is safe
+        unsafe { intrinsics::simd_as(self) }
+    }
+
+    fn from_bits(bits: Self::Bits) -> Self {
+        // Casting usize to pointers is safe
+        unsafe { intrinsics::simd_as(bits) }
+    }
+}
diff --git a/crates/core_simd/src/eq.rs b/crates/core_simd/src/eq.rs
index c7111f720a8..149380746e7 100644
--- a/crates/core_simd/src/eq.rs
+++ b/crates/core_simd/src/eq.rs
@@ -71,3 +71,45 @@ fn simd_ne(self, other: Self) -> Self::Mask {
 }
 
 impl_mask! { i8, i16, i32, i64, isize }
+
+impl<T, const LANES: usize> SimdPartialEq for Simd<*const T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    type Mask = Mask<isize, LANES>;
+
+    #[inline]
+    fn simd_eq(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_eq(self, other)) }
+    }
+
+    #[inline]
+    fn simd_ne(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_ne(self, other)) }
+    }
+}
+
+impl<T, const LANES: usize> SimdPartialEq for Simd<*mut T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    type Mask = Mask<isize, LANES>;
+
+    #[inline]
+    fn simd_eq(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_eq(self, other)) }
+    }
+
+    #[inline]
+    fn simd_ne(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_ne(self, other)) }
+    }
+}
diff --git a/crates/core_simd/src/mod.rs b/crates/core_simd/src/mod.rs
index 9909d639874..ece026a448b 100644
--- a/crates/core_simd/src/mod.rs
+++ b/crates/core_simd/src/mod.rs
@@ -7,6 +7,7 @@
 mod to_bytes;
 
 mod alias;
+mod cast;
 mod elements;
 mod eq;
 mod fmt;
@@ -24,6 +25,7 @@ pub mod simd {
     pub(crate) use crate::core_simd::intrinsics;
 
     pub use crate::core_simd::alias::*;
+    pub use crate::core_simd::cast::*;
     pub use crate::core_simd::elements::*;
     pub use crate::core_simd::eq::*;
     pub use crate::core_simd::lane_count::{LaneCount, SupportedLaneCount};
diff --git a/crates/core_simd/src/ord.rs b/crates/core_simd/src/ord.rs
index 9a87bc2e344..95a1ecaeeda 100644
--- a/crates/core_simd/src/ord.rs
+++ b/crates/core_simd/src/ord.rs
@@ -211,3 +211,117 @@ fn simd_clamp(self, min: Self, max: Self) -> Self {
 }
 
 impl_mask! { i8, i16, i32, i64, isize }
+
+impl<T, const LANES: usize> SimdPartialOrd for Simd<*const T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    #[inline]
+    fn simd_lt(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_lt(self, other)) }
+    }
+
+    #[inline]
+    fn simd_le(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_le(self, other)) }
+    }
+
+    #[inline]
+    fn simd_gt(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_gt(self, other)) }
+    }
+
+    #[inline]
+    fn simd_ge(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_ge(self, other)) }
+    }
+}
+
+impl<T, const LANES: usize> SimdOrd for Simd<*const T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    #[inline]
+    fn simd_max(self, other: Self) -> Self {
+        self.simd_lt(other).select(other, self)
+    }
+
+    #[inline]
+    fn simd_min(self, other: Self) -> Self {
+        self.simd_gt(other).select(other, self)
+    }
+
+    #[inline]
+    fn simd_clamp(self, min: Self, max: Self) -> Self {
+        assert!(
+            min.simd_le(max).all(),
+            "each lane in `min` must be less than or equal to the corresponding lane in `max`",
+        );
+        self.simd_max(min).simd_min(max)
+    }
+}
+
+impl<T, const LANES: usize> SimdPartialOrd for Simd<*mut T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    #[inline]
+    fn simd_lt(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_lt(self, other)) }
+    }
+
+    #[inline]
+    fn simd_le(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_le(self, other)) }
+    }
+
+    #[inline]
+    fn simd_gt(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_gt(self, other)) }
+    }
+
+    #[inline]
+    fn simd_ge(self, other: Self) -> Self::Mask {
+        // Safety: `self` is a vector, and the result of the comparison
+        // is always a valid mask.
+        unsafe { Mask::from_int_unchecked(intrinsics::simd_ge(self, other)) }
+    }
+}
+
+impl<T, const LANES: usize> SimdOrd for Simd<*mut T, LANES>
+where
+    LaneCount<LANES>: SupportedLaneCount,
+{
+    #[inline]
+    fn simd_max(self, other: Self) -> Self {
+        self.simd_lt(other).select(other, self)
+    }
+
+    #[inline]
+    fn simd_min(self, other: Self) -> Self {
+        self.simd_gt(other).select(other, self)
+    }
+
+    #[inline]
+    fn simd_clamp(self, min: Self, max: Self) -> Self {
+        assert!(
+            min.simd_le(max).all(),
+            "each lane in `min` must be less than or equal to the corresponding lane in `max`",
+        );
+        self.simd_max(min).simd_min(max)
+    }
+}
diff --git a/crates/core_simd/src/vector.rs b/crates/core_simd/src/vector.rs
index 7f0e8350cf8..cbc8ced5a84 100644
--- a/crates/core_simd/src/vector.rs
+++ b/crates/core_simd/src/vector.rs
@@ -2,7 +2,7 @@
 pub(crate) mod ptr;
 
 use crate::simd::{
-    intrinsics, LaneCount, Mask, MaskElement, SimdPartialOrd, SupportedLaneCount, Swizzle,
+    intrinsics, LaneCount, Mask, MaskElement, SimdCast, SimdPartialOrd, SupportedLaneCount, Swizzle,
 };
 
 /// A SIMD vector of `LANES` elements of type `T`. `Simd<T, N>` has the same shape as [`[T; N]`](array), but operates like `T`.
@@ -211,7 +211,10 @@ pub const fn from_slice(slice: &[T]) -> Self {
     #[must_use]
     #[inline]
     #[cfg(not(bootstrap))]
-    pub fn cast<U: SimdElement>(self) -> Simd<U, LANES> {
+    pub fn cast<U: SimdElement>(self) -> Simd<U, LANES>
+    where
+        T: SimdCast<U>,
+    {
         // Safety: The input argument is a vector of a valid SIMD element type.
         unsafe { intrinsics::simd_as(self) }
     }
@@ -234,7 +237,7 @@ pub fn cast<U: SimdElement>(self) -> Simd<U, LANES> {
     #[cfg_attr(miri, track_caller)] // even without panics, this helps for Miri backtraces
     pub unsafe fn to_int_unchecked<I>(self) -> Simd<I, LANES>
     where
-        T: core::convert::FloatToInt<I>,
+        T: core::convert::FloatToInt<I> + SimdCast<I>,
         I: SimdElement,
     {
         // Safety: `self` is a vector, and `FloatToInt` ensures the type can be casted to
@@ -739,3 +742,13 @@ impl Sealed for f64 {}
 unsafe impl SimdElement for f64 {
     type Mask = i64;
 }
+
+impl<T> Sealed for *const T {}
+unsafe impl<T> SimdElement for *const T {
+    type Mask = isize;
+}
+
+impl<T> Sealed for *mut T {}
+unsafe impl<T> SimdElement for *mut T {
+    type Mask = isize;
+}