diff --git a/crates/core_simd/src/fmt.rs b/crates/core_simd/src/fmt.rs
index c634e0546bc..16b8f3b95d9 100644
--- a/crates/core_simd/src/fmt.rs
+++ b/crates/core_simd/src/fmt.rs
@@ -30,10 +30,10 @@ debug_wrapper! {
 }
 
 macro_rules! impl_fmt_trait {
-    { $($type:ty => $(($trait:ident, $format:ident)),*;)* } => {
+    { $($type:ident => $(($trait:ident, $format:ident)),*;)* } => {
         $( // repeat type
             $( // repeat trait
-                impl core::fmt::$trait for $type {
+                impl<const LANES: usize> core::fmt::$trait for crate::$type<LANES> {
                     fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
                         $format(self.as_ref(), f)
                     }
@@ -41,7 +41,7 @@ macro_rules! impl_fmt_trait {
             )*
         )*
     };
-    { integers: $($type:ty,)* } => {
+    { integers: $($type:ident,)* } => {
         impl_fmt_trait! {
             $($type =>
               (Debug, format),
@@ -54,7 +54,7 @@ macro_rules! impl_fmt_trait {
             )*
         }
     };
-    { floats: $($type:ty,)* } => {
+    { floats: $($type:ident,)* } => {
         impl_fmt_trait! {
             $($type =>
               (Debug, format),
@@ -63,7 +63,7 @@ macro_rules! impl_fmt_trait {
             )*
         }
     };
-    { masks: $($type:ty,)* } => {
+    { masks: $($type:ident,)* } => {
         impl_fmt_trait! {
             $($type =>
               (Debug, format);
@@ -74,32 +74,12 @@ macro_rules! impl_fmt_trait {
 
 impl_fmt_trait! {
     integers:
-        crate::u8x8,    crate::u8x16,    crate::u8x32,    crate::u8x64,
-        crate::i8x8,    crate::i8x16,    crate::i8x32,    crate::i8x64,
-        crate::u16x4,   crate::u16x8,   crate::u16x16,   crate::u16x32,
-        crate::i16x4,   crate::i16x8,   crate::i16x16,   crate::i16x32,
-        crate::u32x2,   crate::u32x4,   crate::u32x8,   crate::u32x16,
-        crate::i32x2,   crate::i32x4,   crate::i32x8,   crate::i32x16,
-        crate::u64x2,   crate::u64x4,   crate::u64x8,
-        crate::i64x2,   crate::i64x4,   crate::i64x8,
-        crate::u128x2,  crate::u128x4,
-        crate::i128x2,  crate::i128x4,
-        crate::usizex2, crate::usizex4, crate::usizex8,
-        crate::isizex2, crate::isizex4, crate::isizex8,
+        SimdU8, SimdU16, SimdU32, SimdU64, SimdU128,
+        SimdI8, SimdI16, SimdI32, SimdI64, SimdI128,
+        SimdUsize, SimdIsize,
 }
 
 impl_fmt_trait! {
     floats:
-        crate::f32x2, crate::f32x4, crate::f32x8, crate::f32x16,
-        crate::f64x2, crate::f64x4, crate::f64x8,
-}
-
-impl_fmt_trait! {
-    masks:
-        crate::mask8x8,    crate::mask8x16,    crate::mask8x32,    crate::mask8x64,
-        crate::mask16x4,   crate::mask16x8,   crate::mask16x16,   crate::mask16x32,
-        crate::mask32x2,   crate::mask32x4,   crate::mask32x8,   crate::mask32x16,
-        crate::mask64x2,   crate::mask64x4,   crate::mask64x8,
-        crate::mask128x2,  crate::mask128x4,
-        crate::masksizex2, crate::masksizex4, crate::masksizex8,
+        SimdF32, SimdF64,
 }
diff --git a/crates/core_simd/src/intrinsics.rs b/crates/core_simd/src/intrinsics.rs
index b2edc3747ef..51689cd97be 100644
--- a/crates/core_simd/src/intrinsics.rs
+++ b/crates/core_simd/src/intrinsics.rs
@@ -1,7 +1,7 @@
 //! This module contains the LLVM intrinsics bindings that provide the functionality for this
 //! crate.
 //!
-//! The LLVM assembly language is documented here: https://llvm.org/docs/LangRef.html
+//! The LLVM assembly language is documented here: <https://llvm.org/docs/LangRef.html>
 
 /// These intrinsics aren't linked directly from LLVM and are mostly undocumented, however they are
 /// simply lowered to the matching LLVM instructions by the compiler.  The associated instruction
@@ -45,4 +45,11 @@ extern "platform-intrinsic" {
 
     // ceil
     pub(crate) fn simd_ceil<T>(x: T) -> T;
+
+    pub(crate) fn simd_eq<T, U>(x: T, y: T) -> U;
+    pub(crate) fn simd_ne<T, U>(x: T, y: T) -> U;
+    pub(crate) fn simd_lt<T, U>(x: T, y: T) -> U;
+    pub(crate) fn simd_le<T, U>(x: T, y: T) -> U;
+    pub(crate) fn simd_gt<T, U>(x: T, y: T) -> U;
+    pub(crate) fn simd_ge<T, U>(x: T, y: T) -> U;
 }
diff --git a/crates/core_simd/src/lib.rs b/crates/core_simd/src/lib.rs
index cea39e6f3f3..3a3346975c1 100644
--- a/crates/core_simd/src/lib.rs
+++ b/crates/core_simd/src/lib.rs
@@ -1,5 +1,5 @@
 #![no_std]
-#![feature(repr_simd, platform_intrinsics, link_llvm_intrinsics, simd_ffi)]
+#![feature(repr_simd, platform_intrinsics, link_llvm_intrinsics, simd_ffi, min_const_generics)]
 #![warn(missing_docs)]
 //! Portable SIMD module.
 
@@ -9,6 +9,7 @@ mod macros;
 mod fmt;
 mod intrinsics;
 mod ops;
+mod round;
 
 mod masks;
 pub use masks::*;
@@ -43,18 +44,3 @@ mod vectors_f32;
 pub use vectors_f32::*;
 mod vectors_f64;
 pub use vectors_f64::*;
-
-mod vectors_mask8;
-pub use vectors_mask8::*;
-mod vectors_mask16;
-pub use vectors_mask16::*;
-mod vectors_mask32;
-pub use vectors_mask32::*;
-mod vectors_mask64;
-pub use vectors_mask64::*;
-mod vectors_mask128;
-pub use vectors_mask128::*;
-mod vectors_masksize;
-pub use vectors_masksize::*;
-
-mod round;
diff --git a/crates/core_simd/src/macros.rs b/crates/core_simd/src/macros.rs
index f37d13c3ca3..7e4d8514249 100644
--- a/crates/core_simd/src/macros.rs
+++ b/crates/core_simd/src/macros.rs
@@ -26,183 +26,123 @@ macro_rules! from_transmute_x86 {
     }
 }
 
-/// Calls a the macro `$mac` with the provided `$args` followed by `$repeat` repeated the specified
-/// number of times.
-macro_rules! call_repeat {
-    { 1 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)*
-        }
-    };
-    { 2 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)*
-        }
-    };
-    { 4 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-        }
-    };
-    { 8 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-        }
-    };
-    { 16 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-        }
-    };
-    { 32 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-        }
-    };
-    { 64 => $mac:path [$($repeat:tt)*] $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-            $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)* $($repeat)*
-        }
-    };
-}
+/// Implements common traits on the specified vector `$name`, holding multiple `$lanes` of `$type`.
+macro_rules! impl_vector {
+    { $name:ident, $type:ty } => {
+        impl<const LANES: usize> $name<LANES> {
+            /// Construct a SIMD vector by setting all lanes to the given value.
+            pub const fn splat(value: $type) -> Self {
+                Self([value; LANES])
+            }
 
-/// Calls the macro `$mac` with the specified `$args` followed by the specified number of unique
-/// identifiers.
-macro_rules! call_counting_args {
-    { 1 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            value
-        }
-    };
-    { 2 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0 v1
-        }
-    };
-    { 4 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0 v1 v2 v3
-        }
-    };
-    { 8 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0 v1 v2 v3 v4 v5 v6 v7
-        }
-    };
-    { 16 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0 v1 v2 v3 v4 v5 v6 v7 v8 v9 v10 v11 v12 v13 v14 v15
-        }
-    };
-    { 32 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0  v1  v2  v3  v4  v5  v6  v7  v8  v9  v10 v11 v12 v13 v14 v15
-            v16 v17 v18 v19 v20 v21 v22 v23 v24 v25 v26 v27 v28 v29 v30 v31
-        }
-    };
-    { 64 => $mac:path => $($args:tt)* } => {
-        $mac! {
-            $($args)*
-            v0  v1  v2  v3  v4  v5  v6  v7  v8  v9  v10 v11 v12 v13 v14 v15
-            v16 v17 v18 v19 v20 v21 v22 v23 v24 v25 v26 v27 v28 v29 v30 v31
-            v32 v33 v34 v35 v36 v37 v38 v39 v40 v41 v42 v43 v44 v45 v46 v47
-            v48 v49 v50 v51 v52 v53 v54 v55 v56 v57 v58 v59 v60 v61 v62 v63
+            /// Returns a slice containing the entire SIMD vector.
+            pub const fn as_slice(&self) -> &[$type] {
+                &self.0
+            }
+
+            /// Returns a mutable slice containing the entire SIMD vector.
+            pub fn as_mut_slice(&mut self) -> &mut [$type] {
+                &mut self.0
+            }
+
+            /// Converts an array to a SIMD vector.
+            pub const fn from_array(array: [$type; LANES]) -> Self {
+                Self(array)
+            }
+
+            /// Converts a SIMD vector to an array.
+            pub const fn to_array(self) -> [$type; LANES] {
+                // workaround for rust-lang/rust#80108
+                // TODO fix this
+                #[cfg(target_arch = "wasm32")]
+                {
+                    let mut arr = [self.0[0]; LANES];
+                    let mut i = 0;
+                    while i < LANES {
+                        arr[i] = self.0[i];
+                        i += 1;
+                    }
+                    arr
+                }
+
+                #[cfg(not(target_arch = "wasm32"))]
+                {
+                    self.0
+                }
+            }
         }
-    };
-}
 
-/// Implements common traits on the specified vector `$name`, holding multiple `$lanes` of `$type`.
-macro_rules! base_vector_traits {
-    { $name:path => [$type:ty; $lanes:literal] } => {
-        impl Copy for $name {}
+        impl<const LANES: usize> Copy for $name<LANES> {}
 
-        impl Clone for $name {
+        impl<const LANES: usize> Clone for $name<LANES> {
             #[inline]
             fn clone(&self) -> Self {
                 *self
             }
         }
 
-        impl Default for $name {
+        impl<const LANES: usize> Default for $name<LANES> {
             #[inline]
             fn default() -> Self {
                 Self::splat(<$type>::default())
             }
         }
 
-        impl PartialEq for $name {
+        impl<const LANES: usize> PartialEq for $name<LANES> {
             #[inline]
             fn eq(&self, other: &Self) -> bool {
-                AsRef::<[$type]>::as_ref(self) == AsRef::<[$type]>::as_ref(other)
+                // TODO use SIMD equality
+                self.to_array() == other.to_array()
             }
         }
 
-        impl PartialOrd for $name {
+        impl<const LANES: usize> PartialOrd for $name<LANES> {
             #[inline]
             fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
-                AsRef::<[$type]>::as_ref(self).partial_cmp(AsRef::<[$type]>::as_ref(other))
+                // TODO use SIMD equalitya
+                self.to_array().partial_cmp(other.as_ref())
             }
         }
 
         // array references
-        impl AsRef<[$type; $lanes]> for $name {
+        impl<const LANES: usize> AsRef<[$type; LANES]> for $name<LANES> {
             #[inline]
-            fn as_ref(&self) -> &[$type; $lanes] {
-                unsafe { &*(self as *const _ as *const _) }
+            fn as_ref(&self) -> &[$type; LANES] {
+                &self.0
             }
         }
 
-        impl AsMut<[$type; $lanes]> for $name {
+        impl<const LANES: usize> AsMut<[$type; LANES]> for $name<LANES> {
             #[inline]
-            fn as_mut(&mut self) -> &mut [$type; $lanes] {
-                unsafe { &mut *(self as *mut _ as *mut _) }
+            fn as_mut(&mut self) -> &mut [$type; LANES] {
+                &mut self.0
             }
         }
 
         // slice references
-        impl AsRef<[$type]> for $name {
+        impl<const LANES: usize> AsRef<[$type]> for $name<LANES> {
             #[inline]
             fn as_ref(&self) -> &[$type] {
-                AsRef::<[$type; $lanes]>::as_ref(self)
+                &self.0
             }
         }
 
-        impl AsMut<[$type]> for $name {
+        impl<const LANES: usize> AsMut<[$type]> for $name<LANES> {
             #[inline]
             fn as_mut(&mut self) -> &mut [$type] {
-                AsMut::<[$type; $lanes]>::as_mut(self)
+                &mut self.0
             }
         }
 
         // vector/array conversion
-        from_transmute! { unsafe $name => [$type; $lanes] }
+        impl<const LANES: usize> From<[$type; LANES]> for $name<LANES> {
+            fn from(array: [$type; LANES]) -> Self {
+                Self(array)
+            }
+        }
 
         // splat
-        impl From<$type> for $name {
+        impl<const LANES: usize> From<$type> for $name<LANES> {
             #[inline]
             fn from(value: $type) -> Self {
                 Self::splat(value)
@@ -212,154 +152,63 @@ macro_rules! base_vector_traits {
 }
 
 /// Implements additional integer traits (Eq, Ord, Hash) on the specified vector `$name`, holding multiple `$lanes` of `$type`.
-macro_rules! integer_vector_traits {
-    { $name:path => [$type:ty; $lanes:literal] } => {
-        impl Eq for $name {}
+macro_rules! impl_integer_vector {
+    { $name:ident, $type:ty } => {
+        impl_vector! { $name, $type }
 
-        impl Ord for $name {
+        impl<const LANES: usize> Eq for $name<LANES> {}
+
+        impl<const LANES: usize> Ord for $name<LANES> {
             #[inline]
             fn cmp(&self, other: &Self) -> core::cmp::Ordering {
-                AsRef::<[$type]>::as_ref(self).cmp(AsRef::<[$type]>::as_ref(other))
+                // TODO use SIMD cmp
+                self.to_array().cmp(other.as_ref())
             }
         }
 
-        impl core::hash::Hash for $name {
+        impl<const LANES: usize> core::hash::Hash for $name<LANES> {
             #[inline]
             fn hash<H>(&self, state: &mut H)
             where
                 H: core::hash::Hasher
             {
-                AsRef::<[$type]>::as_ref(self).hash(state)
+                self.as_slice().hash(state)
             }
         }
     }
 }
 
-/// Defines a vector `$name` containing multiple `$lanes` of `$type`.
-macro_rules! define_vector {
-    { $(#[$attr:meta])* struct $name:ident([$type:ty; $lanes:tt]); } => {
-        call_repeat! { $lanes => define_vector [$type] def $(#[$attr])* | $name | }
-
-        impl $name {
-            call_repeat! { $lanes => define_vector [$type] splat $type | }
-            call_counting_args! { $lanes => define_vector => new $type | }
-        }
-
-        base_vector_traits! { $name => [$type; $lanes] }
-    };
-    { def $(#[$attr:meta])* | $name:ident | $($itype:ty)* } => {
-        $(#[$attr])*
-        #[allow(non_camel_case_types)]
-        #[repr(simd)]
-        pub struct $name($($itype),*);
-    };
-    { splat $type:ty | $($itype:ty)* } => {
-        /// Construct a vector by setting all lanes to the given value.
-        #[inline]
-        pub const fn splat(value: $type) -> Self {
-            Self($(value as $itype),*)
-        }
-    };
-    { new $type:ty | $($var:ident)* } => {
-        /// Construct a vector by setting each lane to the given values.
-        #[allow(clippy::too_many_arguments)]
-        #[inline]
-        pub const fn new($($var: $type),*) -> Self {
-            Self($($var),*)
-        }
-    }
-}
-
 /// Implements inherent methods for a float vector `$name` containing multiple
 /// `$lanes` of float `$type`, which uses `$bits_ty` as its binary
 /// representation. Called from `define_float_vector!`.
 macro_rules! impl_float_vector {
-    { $name:path => [$type:ty; $lanes:literal]; bits $bits_ty:ty; } => {
-        impl $name {
+    { $name:ident, $type:ty, $bits_ty:ident } => {
+        impl_vector! { $name, $type }
+
+        impl<const LANES: usize> $name<LANES> {
             /// Raw transmutation to an unsigned integer vector type with the
             /// same size and number of lanes.
             #[inline]
-            pub fn to_bits(self) -> $bits_ty {
-                unsafe { core::mem::transmute(self) }
+            pub fn to_bits(self) -> crate::$bits_ty<LANES> {
+                assert_eq!(core::mem::size_of::<Self>(), core::mem::size_of::<crate::$bits_ty<LANES>>());
+                unsafe { core::mem::transmute_copy(&self) }
             }
 
             /// Raw transmutation from an unsigned integer vector type with the
             /// same size and number of lanes.
             #[inline]
-            pub fn from_bits(bits: $bits_ty) -> Self {
-                unsafe { core::mem::transmute(bits) }
+            pub fn from_bits(bits: crate::$bits_ty<LANES>) -> Self {
+                assert_eq!(core::mem::size_of::<Self>(), core::mem::size_of::<crate::$bits_ty<LANES>>());
+                unsafe { core::mem::transmute_copy(&bits) }
             }
 
             /// Produces a vector where every lane has the absolute value of the
             /// equivalently-indexed lane in `self`.
             #[inline]
             pub fn abs(self) -> Self {
-                let no_sign = <$bits_ty>::splat(!0 >> 1);
+                let no_sign = crate::$bits_ty::splat(!0 >> 1);
                 Self::from_bits(self.to_bits() & no_sign)
             }
         }
     };
 }
-
-/// Defines a float vector `$name` containing multiple `$lanes` of float
-/// `$type`, which uses `$bits_ty` as its binary representation.
-macro_rules! define_float_vector {
-    { $(#[$attr:meta])* struct $name:ident([$type:ty; $lanes:tt]); bits $bits_ty:ty; } => {
-        define_vector! {
-            $(#[$attr])*
-            struct $name([$type; $lanes]);
-        }
-
-        impl_float_vector! { $name => [$type; $lanes]; bits $bits_ty; }
-    }
-}
-
-
-/// Defines an integer vector `$name` containing multiple `$lanes` of integer `$type`.
-macro_rules! define_integer_vector {
-    { $(#[$attr:meta])* struct $name:ident([$type:ty; $lanes:tt]); } => {
-        define_vector! {
-            $(#[$attr])*
-            struct $name([$type; $lanes]);
-        }
-
-        integer_vector_traits! { $name => [$type; $lanes] }
-    }
-}
-
-/// Defines a mask vector `$name` containing multiple `$lanes` of `$type`, represented by the
-/// underlying type `$impl_type`.
-macro_rules! define_mask_vector {
-    { $(#[$attr:meta])* struct $name:ident([$impl_type:ty as $type:ty; $lanes:tt]); } => {
-        call_repeat! { $lanes => define_mask_vector [$impl_type] def $(#[$attr])* | $name | }
-
-        impl $name {
-            call_repeat! { $lanes => define_mask_vector [$impl_type] splat $type | }
-            call_counting_args! { $lanes => define_mask_vector => new $type | }
-        }
-
-        base_vector_traits! { $name => [$type; $lanes] }
-        integer_vector_traits! { $name => [$type; $lanes] }
-    };
-    { def $(#[$attr:meta])* | $name:ident | $($itype:ty)* } => {
-        $(#[$attr])*
-        #[allow(non_camel_case_types)]
-        #[repr(simd)]
-        pub struct $name($($itype),*);
-    };
-    { splat $type:ty | $($itype:ty)* } => {
-        /// Construct a vector by setting all lanes to the given value.
-        #[inline]
-        pub const fn splat(value: $type) -> Self {
-            Self($(value.0 as $itype),*)
-        }
-    };
-    { new $type:ty | $($var:ident)* } => {
-        /// Construct a vector by setting each lane to the given values.
-        #[allow(clippy::too_many_arguments)]
-        #[inline]
-        pub const fn new($($var: $type),*) -> Self {
-            Self($($var.0),*)
-        }
-    }
-}
diff --git a/crates/core_simd/src/masks.rs b/crates/core_simd/src/masks.rs
deleted file mode 100644
index cba76b6a2a3..00000000000
--- a/crates/core_simd/src/masks.rs
+++ /dev/null
@@ -1,124 +0,0 @@
-/// The error type returned when converting an integer to a mask fails.
-#[derive(Debug, Copy, Clone, PartialEq, Eq)]
-pub struct TryFromMaskError(());
-
-impl core::fmt::Display for TryFromMaskError {
-    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-        write!(f, "mask must have all bits set or unset")
-    }
-}
-
-macro_rules! define_mask {
-    { $(#[$attr:meta])* struct $name:ident($type:ty); } => {
-        $(#[$attr])*
-        #[allow(non_camel_case_types)]
-        #[derive(Copy, Clone, Default, PartialEq, PartialOrd, Eq, Ord, Hash)]
-        #[repr(transparent)]
-        pub struct $name(pub(crate) $type);
-
-        impl $name {
-            /// Construct a mask from the given value.
-            pub const fn new(value: bool) -> Self {
-                if value {
-                    Self(!0)
-                } else {
-                    Self(0)
-                }
-            }
-
-            /// Test if the mask is set.
-            pub const fn test(&self) -> bool {
-                self.0 != 0
-            }
-        }
-
-        impl core::convert::From<bool> for $name {
-            fn from(value: bool) -> Self {
-                Self::new(value)
-            }
-        }
-
-        impl core::convert::From<$name> for bool {
-            fn from(mask: $name) -> Self {
-                mask.test()
-            }
-        }
-
-        impl core::convert::TryFrom<$type> for $name {
-            type Error = TryFromMaskError;
-            fn try_from(value: $type) -> Result<Self, Self::Error> {
-                if value == 0 || !value == 0 {
-                    Ok(Self(value))
-                } else {
-                    Err(TryFromMaskError(()))
-                }
-            }
-        }
-
-        impl core::convert::From<$name> for $type {
-            fn from(value: $name) -> Self {
-                value.0
-            }
-        }
-
-        impl core::fmt::Debug for $name {
-            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-                self.test().fmt(f)
-            }
-        }
-
-        impl core::fmt::Binary for $name {
-            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-                <$type as core::fmt::Binary>::fmt(&self.0, f)
-            }
-        }
-
-        impl core::fmt::Octal for $name {
-            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-                <$type as core::fmt::Octal>::fmt(&self.0, f)
-            }
-        }
-
-        impl core::fmt::LowerHex for $name {
-            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-                <$type as core::fmt::LowerHex>::fmt(&self.0, f)
-            }
-        }
-
-        impl core::fmt::UpperHex for $name {
-            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
-                <$type as core::fmt::UpperHex>::fmt(&self.0, f)
-            }
-        }
-    }
-}
-
-define_mask! {
-    /// 8-bit mask
-    struct mask8(i8);
-}
-
-define_mask! {
-    /// 16-bit mask
-    struct mask16(i16);
-}
-
-define_mask! {
-    /// 32-bit mask
-    struct mask32(i32);
-}
-
-define_mask! {
-    /// 64-bit mask
-    struct mask64(i64);
-}
-
-define_mask! {
-    /// 128-bit mask
-    struct mask128(i128);
-}
-
-define_mask! {
-    /// `isize`-wide mask
-    struct masksize(isize);
-}
diff --git a/crates/core_simd/src/masks/full_masks/mod.rs b/crates/core_simd/src/masks/full_masks/mod.rs
new file mode 100644
index 00000000000..eac5178512e
--- /dev/null
+++ b/crates/core_simd/src/masks/full_masks/mod.rs
@@ -0,0 +1,205 @@
+//! Masks that take up full SIMD vector registers.
+
+/// The error type returned when converting an integer to a mask fails.
+#[derive(Debug, Copy, Clone, PartialEq, Eq)]
+pub struct TryFromMaskError(());
+
+impl core::fmt::Display for TryFromMaskError {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        write!(f, "mask vector must have all bits set or unset in each lane")
+    }
+}
+
+macro_rules! define_mask {
+    { $(#[$attr:meta])* struct $name:ident<const $lanes:ident: usize>($type:ty); } => {
+        $(#[$attr])*
+        #[derive(Copy, Clone, Default, PartialEq, PartialOrd, Eq, Ord, Hash)]
+        #[repr(transparent)]
+        pub struct $name<const $lanes: usize>($type);
+
+        delegate_ops_to_inner! { $name }
+
+        impl<const $lanes: usize> $name<$lanes> {
+            /// Construct a mask by setting all lanes to the given value.
+            pub fn splat(value: bool) -> Self {
+                Self(<$type>::splat(
+                    if value {
+                        -1
+                    } else {
+                        0
+                    }
+                ))
+            }
+
+            /// Tests the value of the specified lane.
+            ///
+            /// # Panics
+            /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
+            #[inline]
+            pub fn test(&self, lane: usize) -> bool {
+                self.0[lane] == -1
+            }
+
+            /// Sets the value of the specified lane.
+            ///
+            /// # Panics
+            /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
+            #[inline]
+            pub fn set(&mut self, lane: usize, value: bool) {
+                self.0[lane] = if value {
+                    -1
+                } else {
+                    0
+                }
+            }
+        }
+
+        impl<const $lanes: usize> core::convert::From<bool> for $name<$lanes> {
+            fn from(value: bool) -> Self {
+                Self::splat(value)
+            }
+        }
+
+        impl<const $lanes: usize> core::convert::TryFrom<$type> for $name<$lanes> {
+            type Error = TryFromMaskError;
+            fn try_from(value: $type) -> Result<Self, Self::Error> {
+                if value.as_slice().iter().all(|x| *x == 0 || *x == -1) {
+                    Ok(Self(value))
+                } else {
+                    Err(TryFromMaskError(()))
+                }
+            }
+        }
+
+        impl<const $lanes: usize> core::convert::From<$name<$lanes>> for $type {
+            fn from(value: $name<$lanes>) -> Self {
+                value.0
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::Debug for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                f.debug_list()
+                    .entries((0..LANES).map(|lane| self.test(lane)))
+                    .finish()
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::Binary for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                core::fmt::Binary::fmt(&self.0, f)
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::Octal for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                core::fmt::Octal::fmt(&self.0, f)
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::LowerHex for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                core::fmt::LowerHex::fmt(&self.0, f)
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::UpperHex for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                core::fmt::UpperHex::fmt(&self.0, f)
+            }
+        }
+    }
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdI8](crate::SimdI8), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdI8Mask<const LANES: usize>(crate::SimdI8<LANES>);
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdI16](crate::SimdI16), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdI16Mask<const LANES: usize>(crate::SimdI16<LANES>);
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdI32](crate::SimdI32), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdI32Mask<const LANES: usize>(crate::SimdI32<LANES>);
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdI64](crate::SimdI64), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdI64Mask<const LANES: usize>(crate::SimdI64<LANES>);
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdI128](crate::SimdI128), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdI128Mask<const LANES: usize>(crate::SimdI64<LANES>);
+}
+
+define_mask! {
+    /// A mask equivalent to [SimdIsize](crate::SimdIsize), where all bits in the lane must be either set
+    /// or unset.
+    struct SimdIsizeMask<const LANES: usize>(crate::SimdI64<LANES>);
+}
+
+macro_rules! implement_mask_ext {
+    { $($vector:ident => $mask:ident,)* } => {
+        $(
+            impl<const LANES: usize> crate::masks::MaskExt<$mask<LANES>> for crate::$vector<LANES> {
+                #[inline]
+                fn lanes_eq(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_eq(self, other) }
+                }
+
+                #[inline]
+                fn lanes_ne(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_ne(self, other) }
+                }
+
+                #[inline]
+                fn lanes_lt(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_lt(self, other) }
+                }
+
+                #[inline]
+                fn lanes_gt(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_gt(self, other) }
+                }
+
+                #[inline]
+                fn lanes_le(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_le(self, other) }
+                }
+
+                #[inline]
+                fn lanes_ge(&self, other: &Self) -> $mask<LANES> {
+                    unsafe { crate::intrinsics::simd_ge(self, other) }
+                }
+            }
+        )*
+    }
+}
+
+implement_mask_ext! {
+    SimdI8 => SimdI8Mask,
+    SimdI16 => SimdI16Mask,
+    SimdI32 => SimdI32Mask,
+    SimdI64 => SimdI64Mask,
+    SimdI128 => SimdI128Mask,
+    SimdIsize => SimdIsizeMask,
+
+    SimdU8 => SimdI8Mask,
+    SimdU16 => SimdI16Mask,
+    SimdU32 => SimdI32Mask,
+    SimdU64 => SimdI64Mask,
+    SimdU128 => SimdI128Mask,
+    SimdUsize => SimdIsizeMask,
+
+    SimdF32 => SimdI32Mask,
+    SimdF64 => SimdI64Mask,
+}
diff --git a/crates/core_simd/src/masks/mod.rs b/crates/core_simd/src/masks/mod.rs
new file mode 100644
index 00000000000..a075040affe
--- /dev/null
+++ b/crates/core_simd/src/masks/mod.rs
@@ -0,0 +1,391 @@
+//! Types and traits associated with masking lanes of vectors.
+#![allow(non_camel_case_types)]
+
+/// Implements bitwise ops on mask types by delegating the operators to the inner type.
+macro_rules! delegate_ops_to_inner {
+    { $name:ident } => {
+        impl<const LANES: usize> core::ops::BitAnd for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitand(self, rhs: Self) -> Self {
+                Self(self.0 & rhs.0)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitAnd<bool> for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitand(self, rhs: bool) -> Self {
+                self & Self::splat(rhs)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitAnd<$name<LANES>> for bool {
+            type Output = $name<LANES>;
+            #[inline]
+            fn bitand(self, rhs: $name<LANES>) -> $name<LANES> {
+                $name::<LANES>::splat(self) & rhs
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitOr for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitor(self, rhs: Self) -> Self {
+                Self(self.0 | rhs.0)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitOr<bool> for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitor(self, rhs: bool) -> Self {
+                self | Self::splat(rhs)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitOr<$name<LANES>> for bool {
+            type Output = $name<LANES>;
+            #[inline]
+            fn bitor(self, rhs: $name<LANES>) -> $name<LANES> {
+                $name::<LANES>::splat(self) | rhs
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitXor for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitxor(self, rhs: Self) -> Self::Output {
+                Self(self.0 ^ rhs.0)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitXor<bool> for $name<LANES> {
+            type Output = Self;
+            #[inline]
+            fn bitxor(self, rhs: bool) -> Self::Output {
+                self ^ Self::splat(rhs)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitXor<$name<LANES>> for bool {
+            type Output = $name<LANES>;
+            #[inline]
+            fn bitxor(self, rhs: $name<LANES>) -> Self::Output {
+                $name::<LANES>::splat(self) ^ rhs
+            }
+        }
+
+        impl<const LANES: usize> core::ops::Not for $name<LANES> {
+            type Output = $name<LANES>;
+            #[inline]
+            fn not(self) -> Self::Output {
+                Self(!self.0)
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitAndAssign for $name<LANES> {
+            #[inline]
+            fn bitand_assign(&mut self, rhs: Self) {
+                self.0 &= rhs.0;
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitAndAssign<bool> for $name<LANES> {
+            #[inline]
+            fn bitand_assign(&mut self, rhs: bool) {
+                *self &= Self::splat(rhs);
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitOrAssign for $name<LANES> {
+            #[inline]
+            fn bitor_assign(&mut self, rhs: Self) {
+                self.0 |= rhs.0;
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitOrAssign<bool> for $name<LANES> {
+            #[inline]
+            fn bitor_assign(&mut self, rhs: bool) {
+                *self |= Self::splat(rhs);
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitXorAssign for $name<LANES> {
+            #[inline]
+            fn bitxor_assign(&mut self, rhs: Self) {
+                self.0 ^= rhs.0;
+            }
+        }
+
+        impl<const LANES: usize> core::ops::BitXorAssign<bool> for $name<LANES> {
+            #[inline]
+            fn bitxor_assign(&mut self, rhs: bool) {
+                *self ^= Self::splat(rhs);
+            }
+        }
+    }
+}
+
+pub mod full_masks;
+
+macro_rules! define_opaque_mask {
+    {
+        $(#[$attr:meta])*
+        struct $name:ident<const $lanes:ident: usize>($inner_ty:ty);
+    } => {
+        $(#[$attr])*
+        #[allow(non_camel_case_types)]
+        pub struct $name<const $lanes: usize>($inner_ty);
+
+        delegate_ops_to_inner! { $name }
+
+        impl<const $lanes: usize> $name<$lanes> {
+            /// Construct a mask by setting all lanes to the given value.
+            pub fn splat(value: bool) -> Self {
+                Self(<$inner_ty>::splat(value))
+            }
+
+            /// Tests the value of the specified lane.
+            ///
+            /// # Panics
+            /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
+            #[inline]
+            pub fn test(&self, lane: usize) -> bool {
+                self.0.test(lane)
+            }
+
+            /// Sets the value of the specified lane.
+            ///
+            /// # Panics
+            /// Panics if `lane` is greater than or equal to the number of lanes in the vector.
+            #[inline]
+            pub fn set(&mut self, lane: usize, value: bool) {
+                self.0.set(lane, value);
+            }
+        }
+
+        impl<const $lanes: usize> Copy for $name<$lanes> {}
+
+        impl<const $lanes: usize> Clone for $name<$lanes> {
+            #[inline]
+            fn clone(&self) -> Self {
+                *self
+            }
+        }
+
+        impl<const $lanes: usize> Default for $name<$lanes> {
+            #[inline]
+            fn default() -> Self {
+                Self::splat(false)
+            }
+        }
+
+        impl<const $lanes: usize> PartialEq for $name<$lanes> {
+            #[inline]
+            fn eq(&self, other: &Self) -> bool {
+                self.0 == other.0
+            }
+        }
+
+        impl<const $lanes: usize> PartialOrd for $name<$lanes> {
+            #[inline]
+            fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
+                self.0.partial_cmp(&other.0)
+            }
+        }
+
+        impl<const $lanes: usize> core::fmt::Debug for $name<$lanes> {
+            fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+                core::fmt::Debug::fmt(&self.0, f)
+            }
+        }
+    };
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` 8-bit elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct Mask8<const LANES: usize>(full_masks::SimdI8Mask<LANES>);
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` 16-bit elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct Mask16<const LANES: usize>(full_masks::SimdI16Mask<LANES>);
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` 32-bit elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct Mask32<const LANES: usize>(full_masks::SimdI32Mask<LANES>);
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` 64-bit elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct Mask64<const LANES: usize>(full_masks::SimdI64Mask<LANES>);
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` 128-bit elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct Mask128<const LANES: usize>(full_masks::SimdI128Mask<LANES>);
+}
+
+define_opaque_mask! {
+    /// Mask for vectors with `LANES` pointer-width elements.
+    ///
+    /// The layout of this type is unspecified.
+    struct MaskSize<const LANES: usize>(full_masks::SimdIsizeMask<LANES>);
+}
+
+/// Mask-related operations using a particular mask layout.
+pub trait MaskExt<Mask> {
+    /// Test if each lane is equal to the corresponding lane in `other`.
+    fn lanes_eq(&self, other: &Self) -> Mask;
+
+    /// Test if each lane is not equal to the corresponding lane in `other`.
+    fn lanes_ne(&self, other: &Self) -> Mask;
+
+    /// Test if each lane is less than the corresponding lane in `other`.
+    fn lanes_lt(&self, other: &Self) -> Mask;
+
+    /// Test if each lane is greater than the corresponding lane in `other`.
+    fn lanes_gt(&self, other: &Self) -> Mask;
+
+    /// Test if each lane is less than or equal to the corresponding lane in `other`.
+    fn lanes_le(&self, other: &Self) -> Mask;
+
+    /// Test if each lane is greater than or equal to the corresponding lane in `other`.
+    fn lanes_ge(&self, other: &Self) -> Mask;
+}
+
+macro_rules! implement_mask_ops {
+    { $($vector:ident => $mask:ident,)* } => {
+        $(
+            impl<const LANES: usize> crate::$vector<LANES> {
+                /// Test if each lane is equal to the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_eq(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_eq(self, other))
+                }
+
+                /// Test if each lane is not equal to the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_ne(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_ne(self, other))
+                }
+
+                /// Test if each lane is less than the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_lt(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_lt(self, other))
+                }
+
+                /// Test if each lane is greater than the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_gt(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_gt(self, other))
+                }
+
+                /// Test if each lane is less than or equal to the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_le(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_le(self, other))
+                }
+
+                /// Test if each lane is greater than or equal to the corresponding lane in `other`.
+                #[inline]
+                pub fn lanes_ge(&self, other: &Self) -> $mask<LANES> {
+                    $mask(MaskExt::lanes_ge(self, other))
+                }
+            }
+        )*
+    }
+}
+
+implement_mask_ops! {
+    SimdI8 => Mask8,
+    SimdI16 => Mask16,
+    SimdI32 => Mask32,
+    SimdI64 => Mask64,
+    SimdI128 => Mask128,
+    SimdIsize => MaskSize,
+
+    SimdU8 => Mask8,
+    SimdU16 => Mask16,
+    SimdU32 => Mask32,
+    SimdU64 => Mask64,
+    SimdU128 => Mask128,
+    SimdUsize => MaskSize,
+
+    SimdF32 => Mask32,
+    SimdF64 => Mask64,
+}
+
+/// Vector of eight 8-bit masks
+pub type mask8x8 = Mask8<8>;
+
+/// Vector of 16 8-bit masks
+pub type mask8x16 = Mask8<16>;
+
+/// Vector of 32 8-bit masks
+pub type mask8x32 = Mask8<32>;
+
+/// Vector of 16 8-bit masks
+pub type mask8x64 = Mask8<64>;
+
+/// Vector of four 16-bit masks
+pub type mask16x4 = Mask16<4>;
+
+/// Vector of eight 16-bit masks
+pub type mask16x8 = Mask16<8>;
+
+/// Vector of 16 16-bit masks
+pub type mask16x16 = Mask16<16>;
+
+/// Vector of 32 16-bit masks
+pub type mask16x32 = Mask32<32>;
+
+/// Vector of two 32-bit masks
+pub type mask32x2 = Mask32<2>;
+
+/// Vector of four 32-bit masks
+pub type mask32x4 = Mask32<4>;
+
+/// Vector of eight 32-bit masks
+pub type mask32x8 = Mask32<8>;
+
+/// Vector of 16 32-bit masks
+pub type mask32x16 = Mask32<16>;
+
+/// Vector of two 64-bit masks
+pub type mask64x2 = Mask64<2>;
+
+/// Vector of four 64-bit masks
+pub type mask64x4 = Mask64<4>;
+
+/// Vector of eight 64-bit masks
+pub type mask64x8 = Mask64<8>;
+
+/// Vector of two 128-bit masks
+pub type mask128x2 = Mask128<2>;
+
+/// Vector of four 128-bit masks
+pub type mask128x4 = Mask128<4>;
+
+/// Vector of two pointer-width masks
+pub type masksizex2 = MaskSize<2>;
+
+/// Vector of four pointer-width masks
+pub type masksizex4 = MaskSize<4>;
+
+/// Vector of eight pointer-width masks
+pub type masksizex8 = MaskSize<8>;
diff --git a/crates/core_simd/src/ops.rs b/crates/core_simd/src/ops.rs
index 5a186649821..a21e9ab5bf1 100644
--- a/crates/core_simd/src/ops.rs
+++ b/crates/core_simd/src/ops.rs
@@ -12,21 +12,21 @@ where
 macro_rules! impl_ref_ops {
     // binary op
     {
-        impl core::ops::$trait:ident<$rhs:ty> for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty {
             type Output = $output:ty;
 
             $(#[$attrs:meta])*
             fn $fn:ident($self_tok:ident, $rhs_arg:ident: $rhs_arg_ty:ty) -> Self::Output $body:tt
         }
     } => {
-        impl core::ops::$trait<$rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type {
             type Output = $output;
 
             $(#[$attrs])*
             fn $fn($self_tok, $rhs_arg: $rhs_arg_ty) -> Self::Output $body
         }
 
-        impl core::ops::$trait<&'_ $rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -35,7 +35,7 @@ macro_rules! impl_ref_ops {
             }
         }
 
-        impl core::ops::$trait<$rhs> for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for &'_ $type {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -44,7 +44,7 @@ macro_rules! impl_ref_ops {
             }
         }
 
-        impl core::ops::$trait<&'_ $rhs> for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for &'_ $type {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -56,17 +56,17 @@ macro_rules! impl_ref_ops {
 
     // binary assignment op
     {
-        impl core::ops::$trait:ident<$rhs:ty> for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty {
             $(#[$attrs:meta])*
             fn $fn:ident(&mut $self_tok:ident, $rhs_arg:ident: $rhs_arg_ty:ty) $body:tt
         }
     } => {
-        impl core::ops::$trait<$rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type {
             $(#[$attrs])*
             fn $fn(&mut $self_tok, $rhs_arg: $rhs_arg_ty) $body
         }
 
-        impl core::ops::$trait<&'_ $rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type {
             $(#[$attrs])*
             fn $fn(&mut $self_tok, $rhs_arg: &$rhs_arg_ty) {
                 core::ops::$trait::$fn($self_tok, *$rhs_arg)
@@ -76,17 +76,17 @@ macro_rules! impl_ref_ops {
 
     // unary op
     {
-        impl core::ops::$trait:ident for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident for $type:ty {
             type Output = $output:ty;
             fn $fn:ident($self_tok:ident) -> Self::Output $body:tt
         }
     } => {
-        impl core::ops::$trait for $type {
+        impl<const $lanes: usize> core::ops::$trait for $type {
             type Output = $output;
             fn $fn($self_tok) -> Self::Output $body
         }
 
-        impl core::ops::$trait for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait for &'_ $type {
             type Output = <$type as core::ops::$trait>::Output;
             fn $fn($self_tok) -> Self::Output {
                 core::ops::$trait::$fn(*$self_tok)
@@ -95,144 +95,76 @@ macro_rules! impl_ref_ops {
     }
 }
 
-/// Implements op traits for masks
-macro_rules! impl_mask_ops {
-    { $($mask:ty),* } => {
-        $(
-            impl_ref_ops! {
-                impl core::ops::BitAnd<$mask> for $mask {
-                    type Output = Self;
-                    fn bitand(self, rhs: Self) -> Self::Output {
-                        Self(self.0 & rhs.0)
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::BitAndAssign<$mask> for $mask {
-                    fn bitand_assign(&mut self, rhs: Self) {
-                        *self = *self & rhs;
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::BitOr<$mask> for $mask {
-                    type Output = Self;
-                    fn bitor(self, rhs: Self) -> Self::Output {
-                        Self(self.0 | rhs.0)
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::BitOrAssign<$mask> for $mask {
-                    fn bitor_assign(&mut self, rhs: Self) {
-                        *self = *self | rhs;
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::BitXor<$mask> for $mask {
-                    type Output = Self;
-                    fn bitxor(self, rhs: Self) -> Self::Output {
-                        Self(self.0 ^ rhs.0)
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::BitXorAssign<$mask> for $mask {
-                    fn bitxor_assign(&mut self, rhs: Self) {
-                        *self = *self ^ rhs;
-                    }
-                }
-            }
-
-            impl_ref_ops! {
-                impl core::ops::Not for $mask {
-                    type Output = Self;
-                    fn not(self) -> Self::Output {
-                        Self(!self.0)
-                    }
-                }
-            }
-        )*
-    }
-}
-impl_mask_ops! { crate::mask8, crate::mask16, crate::mask32, crate::mask64, crate::mask128, crate::masksize }
-
 /// Automatically implements operators over vectors and scalars for a particular vector.
 macro_rules! impl_op {
-    { impl Add for $type:ty, $scalar:ty } => {
+    { impl Add for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Add::add, AddAssign::add_assign, simd_add }
     };
-    { impl Sub for $type:ty, $scalar:ty } => {
+    { impl Sub for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Sub::sub, SubAssign::sub_assign, simd_sub }
     };
-    { impl Mul for $type:ty, $scalar:ty } => {
+    { impl Mul for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Mul::mul, MulAssign::mul_assign, simd_mul }
     };
-    { impl Div for $type:ty, $scalar:ty } => {
+    { impl Div for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Div::div, DivAssign::div_assign, simd_div }
     };
-    { impl Rem for $type:ty, $scalar:ty } => {
+    { impl Rem for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Rem::rem, RemAssign::rem_assign, simd_rem }
     };
-    { impl Shl for $type:ty, $scalar:ty } => {
+    { impl Shl for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Shl::shl, ShlAssign::shl_assign, simd_shl }
     };
-    { impl Shr for $type:ty, $scalar:ty } => {
+    { impl Shr for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, Shr::shr, ShrAssign::shr_assign, simd_shr }
     };
-    { impl BitAnd for $type:ty, $scalar:ty } => {
+    { impl BitAnd for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, BitAnd::bitand, BitAndAssign::bitand_assign, simd_and }
     };
-    { impl BitOr for $type:ty, $scalar:ty } => {
+    { impl BitOr for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, BitOr::bitor, BitOrAssign::bitor_assign, simd_or }
     };
-    { impl BitXor for $type:ty, $scalar:ty } => {
+    { impl BitXor for $type:ident, $scalar:ty } => {
         impl_op! { @binary $type, $scalar, BitXor::bitxor, BitXorAssign::bitxor_assign, simd_xor }
     };
 
-    { impl Not for $type:ty, $scalar:ty } => {
+    { impl Not for $type:ident, $scalar:ty } => {
         impl_ref_ops! {
-            impl core::ops::Not for $type {
+            impl<const LANES: usize> core::ops::Not for crate::$type<LANES> {
                 type Output = Self;
                 fn not(self) -> Self::Output {
-                    self ^ <$type>::splat(!<$scalar>::default())
+                    self ^ Self::splat(!<$scalar>::default())
                 }
             }
         }
     };
 
-    { impl Neg for $type:ty, $scalar:ty } => {
+    { impl Neg for $type:ident, $scalar:ty } => {
         impl_ref_ops! {
-            impl core::ops::Neg for $type {
+            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES> {
                 type Output = Self;
                 fn neg(self) -> Self::Output {
-                    <$type>::splat(0) - self
+                    Self::splat(0) - self
                 }
             }
         }
     };
 
-    { impl Neg for $type:ty, $scalar:ty, @float } => {
+    { impl Neg for $type:ident, $scalar:ty, @float } => {
         impl_ref_ops! {
-            impl core::ops::Neg for $type {
+            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES> {
                 type Output = Self;
                 fn neg(self) -> Self::Output {
                     // FIXME: Replace this with fneg intrinsic once available.
                     // https://github.com/rust-lang/stdsimd/issues/32
-                    Self::from_bits(<$type>::splat(-0.0).to_bits() ^ self.to_bits())
+                    Self::from_bits(Self::splat(-0.0).to_bits() ^ self.to_bits())
                 }
             }
         }
     };
 
-    { impl Index for $type:ty, $scalar:ty } => {
-        impl<I> core::ops::Index<I> for $type
+    { impl Index for $type:ident, $scalar:ty } => {
+        impl<I, const LANES: usize> core::ops::Index<I> for crate::$type<LANES>
         where
             I: core::slice::SliceIndex<[$scalar]>,
         {
@@ -243,7 +175,7 @@ macro_rules! impl_op {
             }
         }
 
-        impl<I> core::ops::IndexMut<I> for $type
+        impl<I, const LANES: usize> core::ops::IndexMut<I> for crate::$type<LANES>
         where
             I: core::slice::SliceIndex<[$scalar]>,
         {
@@ -255,13 +187,13 @@ macro_rules! impl_op {
     };
 
     // generic binary op with assignment when output is `Self`
-    { @binary $type:ty, $scalar:ty, $trait:ident :: $trait_fn:ident, $assign_trait:ident :: $assign_trait_fn:ident, $intrinsic:ident } => {
+    { @binary $type:ident, $scalar:ty, $trait:ident :: $trait_fn:ident, $assign_trait:ident :: $assign_trait_fn:ident, $intrinsic:ident } => {
         impl_ref_ops! {
-            impl core::ops::$trait<$type> for $type {
-                type Output = $type;
+            impl<const LANES: usize> core::ops::$trait<Self> for crate::$type<LANES> {
+                type Output = Self;
 
                 #[inline]
-                fn $trait_fn(self, rhs: $type) -> Self::Output {
+                fn $trait_fn(self, rhs: Self) -> Self::Output {
                     unsafe {
                         crate::intrinsics::$intrinsic(self, rhs)
                     }
@@ -270,31 +202,31 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl core::ops::$trait<$scalar> for $type {
-                type Output = $type;
+            impl<const LANES: usize> core::ops::$trait<$scalar> for crate::$type<LANES> {
+                type Output = Self;
 
                 #[inline]
                 fn $trait_fn(self, rhs: $scalar) -> Self::Output {
-                    core::ops::$trait::$trait_fn(self, <$type>::splat(rhs))
+                    core::ops::$trait::$trait_fn(self, Self::splat(rhs))
                 }
             }
         }
 
         impl_ref_ops! {
-            impl core::ops::$trait<$type> for $scalar {
-                type Output = $type;
+            impl<const LANES: usize> core::ops::$trait<crate::$type<LANES>> for $scalar {
+                type Output = crate::$type<LANES>;
 
                 #[inline]
-                fn $trait_fn(self, rhs: $type) -> Self::Output {
-                    core::ops::$trait::$trait_fn(<$type>::splat(self), rhs)
+                fn $trait_fn(self, rhs: crate::$type<LANES>) -> Self::Output {
+                    core::ops::$trait::$trait_fn(crate::$type::splat(self), rhs)
                 }
             }
         }
 
         impl_ref_ops! {
-            impl core::ops::$assign_trait<$type> for $type {
+            impl<const LANES: usize> core::ops::$assign_trait<Self> for crate::$type<LANES> {
                 #[inline]
-                fn $assign_trait_fn(&mut self, rhs: $type) {
+                fn $assign_trait_fn(&mut self, rhs: Self) {
                     unsafe {
                         *self = crate::intrinsics::$intrinsic(*self, rhs);
                     }
@@ -303,10 +235,10 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl core::ops::$assign_trait<$scalar> for $type {
+            impl<const LANES: usize> core::ops::$assign_trait<$scalar> for crate::$type<LANES> {
                 #[inline]
                 fn $assign_trait_fn(&mut self, rhs: $scalar) {
-                    core::ops::$assign_trait::$assign_trait_fn(self, <$type>::splat(rhs));
+                    core::ops::$assign_trait::$assign_trait_fn(self, Self::splat(rhs));
                 }
             }
         }
@@ -315,7 +247,7 @@ macro_rules! impl_op {
 
 /// Implements floating-point operators for the provided types.
 macro_rules! impl_float_ops {
-    { $($scalar:ty => $($vector:ty),*;)* } => {
+    { $($scalar:ty => $($vector:ident),*;)* } => {
         $( // scalar
             $( // vector
                 impl_op! { impl Add for $vector, $scalar }
@@ -330,24 +262,9 @@ macro_rules! impl_float_ops {
     };
 }
 
-/// Implements mask operators for the provided types.
-macro_rules! impl_mask_ops {
-    { $($scalar:ty => $($vector:ty),*;)* } => {
-        $( // scalar
-            $( // vector
-                impl_op! { impl BitAnd for $vector, $scalar }
-                impl_op! { impl BitOr  for $vector, $scalar }
-                impl_op! { impl BitXor for $vector, $scalar }
-                impl_op! { impl Not for $vector, $scalar }
-                impl_op! { impl Index for $vector, $scalar }
-            )*
-        )*
-    };
-}
-
 /// Implements unsigned integer operators for the provided types.
 macro_rules! impl_unsigned_int_ops {
-    { $($scalar:ty => $($vector:ty),*;)* } => {
+    { $($scalar:ty => $($vector:ident),*;)* } => {
         $( // scalar
             $( // vector
                 impl_op! { impl Add for $vector, $scalar }
@@ -361,11 +278,11 @@ macro_rules! impl_unsigned_int_ops {
 
                 // Integers panic on divide by 0
                 impl_ref_ops! {
-                    impl core::ops::Div<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::Div<Self> for crate::$vector<LANES> {
                         type Output = Self;
 
                         #[inline]
-                        fn div(self, rhs: $vector) -> Self::Output {
+                        fn div(self, rhs: Self) -> Self::Output {
                             // TODO there is probably a better way of doing this
                             if AsRef::<[$scalar]>::as_ref(&rhs)
                                 .iter()
@@ -379,8 +296,8 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Div<$scalar> for $vector {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Div<$scalar> for crate::$vector<LANES> {
+                        type Output = Self;
 
                         #[inline]
                         fn div(self, rhs: $scalar) -> Self::Output {
@@ -394,18 +311,18 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Div<$vector> for $scalar {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Div<crate::$vector<LANES>> for $scalar {
+                        type Output = crate::$vector<LANES>;
 
                         #[inline]
-                        fn div(self, rhs: $vector) -> Self::Output {
-                            <$vector>::splat(self) / rhs
+                        fn div(self, rhs: crate::$vector<LANES>) -> Self::Output {
+                            crate::$vector::splat(self) / rhs
                         }
                     }
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::DivAssign<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::DivAssign<Self> for crate::$vector<LANES> {
                         #[inline]
                         fn div_assign(&mut self, rhs: Self) {
                             *self = *self / rhs;
@@ -414,7 +331,7 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::DivAssign<$scalar> for $vector {
+                    impl<const LANES: usize> core::ops::DivAssign<$scalar> for crate::$vector<LANES> {
                         #[inline]
                         fn div_assign(&mut self, rhs: $scalar) {
                             *self = *self / rhs;
@@ -424,11 +341,11 @@ macro_rules! impl_unsigned_int_ops {
 
                 // remainder panics on zero divisor
                 impl_ref_ops! {
-                    impl core::ops::Rem<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::Rem<Self> for crate::$vector<LANES> {
                         type Output = Self;
 
                         #[inline]
-                        fn rem(self, rhs: $vector) -> Self::Output {
+                        fn rem(self, rhs: Self) -> Self::Output {
                             // TODO there is probably a better way of doing this
                             if AsRef::<[$scalar]>::as_ref(&rhs)
                                 .iter()
@@ -442,8 +359,8 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Rem<$scalar> for $vector {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Rem<$scalar> for crate::$vector<LANES> {
+                        type Output = Self;
 
                         #[inline]
                         fn rem(self, rhs: $scalar) -> Self::Output {
@@ -457,18 +374,18 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Rem<$vector> for $scalar {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Rem<crate::$vector<LANES>> for $scalar {
+                        type Output = crate::$vector<LANES>;
 
                         #[inline]
-                        fn rem(self, rhs: $vector) -> Self::Output {
-                            <$vector>::splat(self) % rhs
+                        fn rem(self, rhs: crate::$vector<LANES>) -> Self::Output {
+                            crate::$vector::splat(self) % rhs
                         }
                     }
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::RemAssign<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::RemAssign<Self> for crate::$vector<LANES> {
                         #[inline]
                         fn rem_assign(&mut self, rhs: Self) {
                             *self = *self % rhs;
@@ -477,7 +394,7 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::RemAssign<$scalar> for $vector {
+                    impl<const LANES: usize> core::ops::RemAssign<$scalar> for crate::$vector<LANES> {
                         #[inline]
                         fn rem_assign(&mut self, rhs: $scalar) {
                             *self = *self % rhs;
@@ -487,11 +404,11 @@ macro_rules! impl_unsigned_int_ops {
 
                 // shifts panic on overflow
                 impl_ref_ops! {
-                    impl core::ops::Shl<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::Shl<Self> for crate::$vector<LANES> {
                         type Output = Self;
 
                         #[inline]
-                        fn shl(self, rhs: $vector) -> Self::Output {
+                        fn shl(self, rhs: Self) -> Self::Output {
                             // TODO there is probably a better way of doing this
                             if AsRef::<[$scalar]>::as_ref(&rhs)
                                 .iter()
@@ -506,8 +423,8 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Shl<$scalar> for $vector {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Shl<$scalar> for crate::$vector<LANES> {
+                        type Output = Self;
 
                         #[inline]
                         fn shl(self, rhs: $scalar) -> Self::Output {
@@ -522,7 +439,7 @@ macro_rules! impl_unsigned_int_ops {
 
 
                 impl_ref_ops! {
-                    impl core::ops::ShlAssign<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::ShlAssign<Self> for crate::$vector<LANES> {
                         #[inline]
                         fn shl_assign(&mut self, rhs: Self) {
                             *self = *self << rhs;
@@ -531,7 +448,7 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::ShlAssign<$scalar> for $vector {
+                    impl<const LANES: usize> core::ops::ShlAssign<$scalar> for crate::$vector<LANES> {
                         #[inline]
                         fn shl_assign(&mut self, rhs: $scalar) {
                             *self = *self << rhs;
@@ -540,13 +457,13 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Shr<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::Shr<Self> for crate::$vector<LANES> {
                         type Output = Self;
 
                         #[inline]
-                        fn shr(self, rhs: $vector) -> Self::Output {
+                        fn shr(self, rhs: Self) -> Self::Output {
                             // TODO there is probably a better way of doing this
-                            if AsRef::<[$scalar]>::as_ref(&rhs)
+                            if rhs.as_slice()
                                 .iter()
                                 .copied()
                                 .any(invalid_shift_rhs)
@@ -559,8 +476,8 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::Shr<$scalar> for $vector {
-                        type Output = $vector;
+                    impl<const LANES: usize> core::ops::Shr<$scalar> for crate::$vector<LANES> {
+                        type Output = Self;
 
                         #[inline]
                         fn shr(self, rhs: $scalar) -> Self::Output {
@@ -575,7 +492,7 @@ macro_rules! impl_unsigned_int_ops {
 
 
                 impl_ref_ops! {
-                    impl core::ops::ShrAssign<$vector> for $vector {
+                    impl<const LANES: usize> core::ops::ShrAssign<Self> for crate::$vector<LANES> {
                         #[inline]
                         fn shr_assign(&mut self, rhs: Self) {
                             *self = *self >> rhs;
@@ -584,7 +501,7 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl core::ops::ShrAssign<$scalar> for $vector {
+                    impl<const LANES: usize> core::ops::ShrAssign<$scalar> for crate::$vector<LANES> {
                         #[inline]
                         fn shr_assign(&mut self, rhs: $scalar) {
                             *self = *self >> rhs;
@@ -598,7 +515,7 @@ macro_rules! impl_unsigned_int_ops {
 
 /// Implements unsigned integer operators for the provided types.
 macro_rules! impl_signed_int_ops {
-    { $($scalar:ty => $($vector:ty),*;)* } => {
+    { $($scalar:ty => $($vector:ident),*;)* } => {
         impl_unsigned_int_ops! { $($scalar => $($vector),*;)* }
         $( // scalar
             $( // vector
@@ -609,33 +526,24 @@ macro_rules! impl_signed_int_ops {
 }
 
 impl_unsigned_int_ops! {
-    u8    => crate::u8x8,    crate::u8x16,   crate::u8x32,   crate::u8x64;
-    u16   => crate::u16x4,   crate::u16x8,   crate::u16x16,  crate::u16x32;
-    u32   => crate::u32x2,   crate::u32x4,   crate::u32x8,   crate::u32x16;
-    u64   => crate::u64x2,   crate::u64x4,   crate::u64x8;
-    u128  => crate::u128x2,  crate::u128x4;
-    usize => crate::usizex2, crate::usizex4, crate::usizex8;
+    u8 => SimdU8;
+    u16 => SimdU16;
+    u32 => SimdU32;
+    u64 => SimdU64;
+    u128 => SimdU128;
+    usize => SimdUsize;
 }
 
 impl_signed_int_ops! {
-    i8    => crate::i8x8,    crate::i8x16,   crate::i8x32,   crate::i8x64;
-    i16   => crate::i16x4,   crate::i16x8,   crate::i16x16,  crate::i16x32;
-    i32   => crate::i32x2,   crate::i32x4,   crate::i32x8,   crate::i32x16;
-    i64   => crate::i64x2,   crate::i64x4,   crate::i64x8;
-    i128  => crate::i128x2,  crate::i128x4;
-    isize => crate::isizex2, crate::isizex4, crate::isizex8;
+    i8 => SimdI8;
+    i16 => SimdI16;
+    i32 => SimdI32;
+    i64 => SimdI64;
+    i128 => SimdI128;
+    isize => SimdIsize;
 }
 
 impl_float_ops! {
-    f32 => crate::f32x2, crate::f32x4, crate::f32x8, crate::f32x16;
-    f64 => crate::f64x2, crate::f64x4, crate::f64x8;
-}
-
-impl_mask_ops! {
-    crate::mask8    => crate::mask8x8,    crate::mask8x16,   crate::mask8x32,   crate::mask8x64;
-    crate::mask16   => crate::mask16x4,   crate::mask16x8,   crate::mask16x16,  crate::mask16x32;
-    crate::mask32   => crate::mask32x2,   crate::mask32x4,   crate::mask32x8,   crate::mask32x16;
-    crate::mask64   => crate::mask64x2,   crate::mask64x4,   crate::mask64x8;
-    crate::mask128  => crate::mask128x2,  crate::mask128x4;
-    crate::masksize => crate::masksizex2, crate::masksizex4, crate::masksizex8;
+    f32 => SimdF32;
+    f64 => SimdF64;
 }
diff --git a/crates/core_simd/src/round.rs b/crates/core_simd/src/round.rs
index 0529bbe0080..d77bc4e8fa7 100644
--- a/crates/core_simd/src/round.rs
+++ b/crates/core_simd/src/round.rs
@@ -1,88 +1,45 @@
 macro_rules! implement {
     {
-        impl $type:ident {
-            int_type = $int_type:ident
-        }
+        $type:ident, $int_type:ident
     } => {
-        mod $type {
-            impl crate::$type {
-                /// Returns the largest integer less than or equal to each lane.
-                #[must_use = "method returns a new vector and does not mutate the original value"]
-                #[inline]
-                pub fn floor(self) -> Self {
-                    unsafe { crate::intrinsics::simd_floor(self) }
-                }
+        impl<const LANES: usize> crate::$type<LANES> {
+            /// Returns the largest integer less than or equal to each lane.
+            #[must_use = "method returns a new vector and does not mutate the original value"]
+            #[inline]
+            pub fn floor(self) -> Self {
+                unsafe { crate::intrinsics::simd_floor(self) }
+            }
 
-                /// Returns the smallest integer greater than or equal to each lane.
-                #[must_use = "method returns a new vector and does not mutate the original value"]
-                #[inline]
-                pub fn ceil(self) -> Self {
-                    unsafe { crate::intrinsics::simd_ceil(self) }
-                }
+            /// Returns the smallest integer greater than or equal to each lane.
+            #[must_use = "method returns a new vector and does not mutate the original value"]
+            #[inline]
+            pub fn ceil(self) -> Self {
+                unsafe { crate::intrinsics::simd_ceil(self) }
+            }
 
-                /// Rounds toward zero and converts to the same-width integer type, assuming that
-                /// the value is finite and fits in that type.
-                ///
-                /// # Safety
-                /// The value must:
-                ///
-                /// * Not be NaN
-                /// * Not be infinite
-                /// * Be representable in the return type, after truncating off its fractional part
-                #[inline]
-                pub unsafe fn to_int_unchecked(self) -> crate::$int_type {
-                    crate::intrinsics::simd_cast(self)
-                }
+            /// Rounds toward zero and converts to the same-width integer type, assuming that
+            /// the value is finite and fits in that type.
+            ///
+            /// # Safety
+            /// The value must:
+            ///
+            /// * Not be NaN
+            /// * Not be infinite
+            /// * Be representable in the return type, after truncating off its fractional part
+            #[inline]
+            pub unsafe fn to_int_unchecked(self) -> crate::$int_type<LANES> {
+                crate::intrinsics::simd_cast(self)
+            }
 
-                /// Creates a floating-point vector from an integer vector.  Rounds values that are
-                /// not exactly representable.
-                #[inline]
-                pub fn round_from_int(value: crate::$int_type) -> Self {
-                    unsafe { crate::intrinsics::simd_cast(value) }
-                }
+            /// Creates a floating-point vector from an integer vector.  Rounds values that are
+            /// not exactly representable.
+            #[inline]
+            pub fn round_from_int(value: crate::$int_type<LANES>) -> Self {
+                unsafe { crate::intrinsics::simd_cast(value) }
             }
         }
     }
 }
 
-implement! {
-    impl f32x2 {
-        int_type = i32x2
-    }
-}
-
-implement! {
-    impl f32x4 {
-        int_type = i32x4
-    }
-}
-
-implement! {
-    impl f32x8 {
-        int_type = i32x8
-    }
-}
-
-implement! {
-    impl f32x16 {
-        int_type = i32x16
-    }
-}
-
-implement! {
-    impl f64x2 {
-        int_type = i64x2
-    }
-}
-
-implement! {
-    impl f64x4 {
-        int_type = i64x4
-    }
-}
-
-implement! {
-    impl f64x8 {
-        int_type = i64x8
-    }
-}
+implement! { SimdF32, SimdI32 }
+implement! { SimdF64, SimdI64 }
diff --git a/crates/core_simd/src/vectors_f32.rs b/crates/core_simd/src/vectors_f32.rs
index 17b382ee739..0b5d8c6ec49 100644
--- a/crates/core_simd/src/vectors_f32.rs
+++ b/crates/core_simd/src/vectors_f32.rs
@@ -1,29 +1,23 @@
-define_float_vector! {
-    /// Vector of two `f32` values
-    struct f32x2([f32; 2]);
-    bits crate::u32x2;
-}
+#![allow(non_camel_case_types)]
 
-define_float_vector! {
-    /// Vector of four `f32` values
-    struct f32x4([f32; 4]);
-    bits crate::u32x4;
-}
+/// A SIMD vector of containing `LANES` `f32` values.
+#[repr(simd)]
+pub struct SimdF32<const LANES: usize>([f32; LANES]);
 
-define_float_vector! {
-    /// Vector of eight `f32` values
-    struct f32x8([f32; 8]);
-    bits crate::u32x8;
-}
+impl_float_vector! { SimdF32, f32, SimdU32 }
 
-define_float_vector! {
-    /// Vector of 16 `f32` values
-    struct f32x16([f32; 16]);
-    bits crate::u32x16;
-}
+/// Vector of two `f32` values
+pub type f32x2 = SimdF32<2>;
+
+/// Vector of four `f32` values
+pub type f32x4 = SimdF32<4>;
+
+/// Vector of eight `f32` values
+pub type f32x8 = SimdF32<8>;
+
+/// Vector of 16 `f32` values
+pub type f32x16 = SimdF32<16>;
 
 from_transmute_x86! { unsafe f32x4 => __m128 }
 from_transmute_x86! { unsafe f32x8 => __m256 }
 //from_transmute_x86! { unsafe f32x16 => __m512 }
-
-
diff --git a/crates/core_simd/src/vectors_f64.rs b/crates/core_simd/src/vectors_f64.rs
index b41923ca6f1..307f8a4acac 100644
--- a/crates/core_simd/src/vectors_f64.rs
+++ b/crates/core_simd/src/vectors_f64.rs
@@ -1,20 +1,19 @@
-define_float_vector! {
-    /// Vector of two `f64` values
-    struct f64x2([f64; 2]);
-    bits crate::u64x2;
-}
+#![allow(non_camel_case_types)]
 
-define_float_vector! {
-    /// Vector of four `f64` values
-    struct f64x4([f64; 4]);
-    bits crate::u64x4;
-}
+/// A SIMD vector of containing `LANES` `f64` values.
+#[repr(simd)]
+pub struct SimdF64<const LANES: usize>([f64; LANES]);
 
-define_float_vector! {
-    /// Vector of eight `f64` values
-    struct f64x8([f64; 8]);
-    bits crate::u64x8;
-}
+impl_float_vector! { SimdF64, f64, SimdU64 }
+
+/// Vector of two `f64` values
+pub type f64x2 = SimdF64<2>;
+
+/// Vector of four `f64` values
+pub type f64x4 = SimdF64<4>;
+
+/// Vector of eight `f64` values
+pub type f64x8 = SimdF64<8>;
 
 from_transmute_x86! { unsafe f64x2 => __m128d }
 from_transmute_x86! { unsafe f64x4 => __m256d }
diff --git a/crates/core_simd/src/vectors_i128.rs b/crates/core_simd/src/vectors_i128.rs
index 5c8354070e8..16e6162be55 100644
--- a/crates/core_simd/src/vectors_i128.rs
+++ b/crates/core_simd/src/vectors_i128.rs
@@ -1,12 +1,16 @@
-define_integer_vector! {
-    /// Vector of two `i128` values
-    struct i128x2([i128; 2]);
-}
-
-define_integer_vector! {
-    /// Vector of four `i128` values
-    struct i128x4([i128; 4]);
-}
+#![allow(non_camel_case_types)]
+
+/// A SIMD vector of containing `LANES` `i128` values.
+#[repr(simd)]
+pub struct SimdI128<const LANES: usize>([i128; LANES]);
+
+impl_integer_vector! { SimdI128, i128 }
+
+/// Vector of two `i128` values
+pub type i128x2 = SimdI128<2>;
+
+/// Vector of four `i128` values
+pub type i128x4 = SimdI128<4>;
 
 from_transmute_x86! { unsafe i128x2 => __m256i }
 //from_transmute_x86! { unsafe i128x4 => __m512i }
diff --git a/crates/core_simd/src/vectors_i16.rs b/crates/core_simd/src/vectors_i16.rs
index 8aabd136b10..08cc4af2a5e 100644
--- a/crates/core_simd/src/vectors_i16.rs
+++ b/crates/core_simd/src/vectors_i16.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of four `i16` values
-    struct i16x4([i16; 4]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of eight `i16` values
-    struct i16x8([i16; 8]);
-}
+/// A SIMD vector of containing `LANES` `i16` values.
+#[repr(simd)]
+pub struct SimdI16<const LANES: usize>([i16; LANES]);
 
-define_integer_vector! {
-    /// Vector of 16 `i16` values
-    struct i16x16([i16; 16]);
-}
+impl_integer_vector! { SimdI16, i16 }
 
-define_integer_vector! {
-    /// Vector of 32 `i16` values
-    struct i16x32([i16; 32]);
-}
+/// Vector of four `i16` values
+pub type i16x4 = SimdI16<4>;
+
+/// Vector of eight `i16` values
+pub type i16x8 = SimdI16<8>;
+
+/// Vector of 16 `i16` values
+pub type i16x16 = SimdI16<16>;
+
+/// Vector of 32 `i16` values
+pub type i16x32 = SimdI16<32>;
 
 from_transmute_x86! { unsafe i16x8 => __m128i }
 from_transmute_x86! { unsafe i16x16 => __m256i }
diff --git a/crates/core_simd/src/vectors_i32.rs b/crates/core_simd/src/vectors_i32.rs
index 9aa9bc8e9dc..116f2abaeee 100644
--- a/crates/core_simd/src/vectors_i32.rs
+++ b/crates/core_simd/src/vectors_i32.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of two `i32` values
-    struct i32x2([i32; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `i32` values
-    struct i32x4([i32; 4]);
-}
+/// A SIMD vector of containing `LANES` `i32` values.
+#[repr(simd)]
+pub struct SimdI32<const LANES: usize>([i32; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `i32` values
-    struct i32x8([i32; 8]);
-}
+impl_integer_vector! { SimdI32, i32 }
 
-define_integer_vector! {
-    /// Vector of 16 `i32` values
-    struct i32x16([i32; 16]);
-}
+/// Vector of two `i32` values
+pub type i32x2 = SimdI32<2>;
+
+/// Vector of four `i32` values
+pub type i32x4 = SimdI32<4>;
+
+/// Vector of eight `i32` values
+pub type i32x8 = SimdI32<8>;
+
+/// Vector of 16 `i32` values
+pub type i32x16 = SimdI32<16>;
 
 from_transmute_x86! { unsafe i32x4 => __m128i }
 from_transmute_x86! { unsafe i32x8 => __m256i }
diff --git a/crates/core_simd/src/vectors_i64.rs b/crates/core_simd/src/vectors_i64.rs
index ba66aba2095..6a1e2094179 100644
--- a/crates/core_simd/src/vectors_i64.rs
+++ b/crates/core_simd/src/vectors_i64.rs
@@ -1,17 +1,19 @@
-define_integer_vector! {
-    /// Vector of two `i64` values
-    struct i64x2([i64; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `i64` values
-    struct i64x4([i64; 4]);
-}
+/// A SIMD vector of containing `LANES` `i64` values.
+#[repr(simd)]
+pub struct SimdI64<const LANES: usize>([i64; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `i64` values
-    struct i64x8([i64; 8]);
-}
+impl_integer_vector! { SimdI64, i64 }
+
+/// Vector of two `i64` values
+pub type i64x2 = SimdI64<2>;
+
+/// Vector of four `i64` values
+pub type i64x4 = SimdI64<4>;
+
+/// Vector of eight `i64` values
+pub type i64x8 = SimdI64<8>;
 
 from_transmute_x86! { unsafe i64x2 => __m128i }
 from_transmute_x86! { unsafe i64x4 => __m256i }
diff --git a/crates/core_simd/src/vectors_i8.rs b/crates/core_simd/src/vectors_i8.rs
index 3e52d894cc2..0ac5ba9efee 100644
--- a/crates/core_simd/src/vectors_i8.rs
+++ b/crates/core_simd/src/vectors_i8.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of eight `i8` values
-    struct i8x8([i8; 8]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of 16 `i8` values
-    struct i8x16([i8; 16]);
-}
+/// A SIMD vector of containing `LANES` `i8` values.
+#[repr(simd)]
+pub struct SimdI8<const LANES: usize>([i8; LANES]);
 
-define_integer_vector! {
-    /// Vector of 32 `i8` values
-    struct i8x32([i8; 32]);
-}
+impl_integer_vector! { SimdI8, i8 }
 
-define_integer_vector! {
-    /// Vector of 64 `i8` values
-    struct i8x64([i8; 64]);
-}
+/// Vector of eight `i8` values
+pub type i8x8 = SimdI8<8>;
+
+/// Vector of 16 `i8` values
+pub type i8x16 = SimdI8<16>;
+
+/// Vector of 32 `i8` values
+pub type i8x32 = SimdI8<32>;
+
+/// Vector of 64 `i8` values
+pub type i8x64 = SimdI8<64>;
 
 from_transmute_x86! { unsafe i8x16 => __m128i }
 from_transmute_x86! { unsafe i8x32 => __m256i }
diff --git a/crates/core_simd/src/vectors_isize.rs b/crates/core_simd/src/vectors_isize.rs
index 35dac8bcbd4..6856f305092 100644
--- a/crates/core_simd/src/vectors_isize.rs
+++ b/crates/core_simd/src/vectors_isize.rs
@@ -1,17 +1,19 @@
-define_integer_vector! {
-    /// Vector of two `isize` values
-    struct isizex2([isize; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `isize` values
-    struct isizex4([isize; 4]);
-}
+/// A SIMD vector of containing `LANES` `isize` values.
+#[repr(simd)]
+pub struct SimdIsize<const LANES: usize>([isize; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `isize` values
-    struct isizex8([isize; 8]);
-}
+impl_integer_vector! { SimdIsize, isize }
+
+/// Vector of two `isize` values
+pub type isizex2 = SimdIsize<2>;
+
+/// Vector of four `isize` values
+pub type isizex4 = SimdIsize<4>;
+
+/// Vector of eight `isize` values
+pub type isizex8 = SimdIsize<8>;
 
 #[cfg(target_pointer_width = "32")]
 from_transmute_x86! { unsafe isizex4 => __m128i }
diff --git a/crates/core_simd/src/vectors_mask128.rs b/crates/core_simd/src/vectors_mask128.rs
deleted file mode 100644
index adf56a3684b..00000000000
--- a/crates/core_simd/src/vectors_mask128.rs
+++ /dev/null
@@ -1,11 +0,0 @@
-use crate::mask128;
-
-define_mask_vector! {
-    /// Vector of two `mask128` values
-    struct mask128x2([i128 as mask128; 2]);
-}
-
-define_mask_vector! {
-    /// Vector of four `mask128` values
-    struct mask128x4([i128 as mask128; 4]);
-}
diff --git a/crates/core_simd/src/vectors_mask16.rs b/crates/core_simd/src/vectors_mask16.rs
deleted file mode 100644
index 406d7255a11..00000000000
--- a/crates/core_simd/src/vectors_mask16.rs
+++ /dev/null
@@ -1,21 +0,0 @@
-use crate::mask16;
-
-define_mask_vector! {
-    /// Vector of four `mask16` values
-    struct mask16x4([i16 as mask16; 4]);
-}
-
-define_mask_vector! {
-    /// Vector of eight `mask16` values
-    struct mask16x8([i16 as mask16; 8]);
-}
-
-define_mask_vector! {
-    /// Vector of 16 `mask16` values
-    struct mask16x16([i16 as mask16; 16]);
-}
-
-define_mask_vector! {
-    /// Vector of 32 `mask16` values
-    struct mask16x32([i16 as mask16; 32]);
-}
diff --git a/crates/core_simd/src/vectors_mask32.rs b/crates/core_simd/src/vectors_mask32.rs
deleted file mode 100644
index fad191421f3..00000000000
--- a/crates/core_simd/src/vectors_mask32.rs
+++ /dev/null
@@ -1,21 +0,0 @@
-use crate::mask32;
-
-define_mask_vector! {
-    /// Vector of two `mask32` values
-    struct mask32x2([i32 as mask32; 2]);
-}
-
-define_mask_vector! {
-    /// Vector of four `mask32` values
-    struct mask32x4([i32 as mask32; 4]);
-}
-
-define_mask_vector! {
-    /// Vector of eight `mask32` values
-    struct mask32x8([i32 as mask32; 8]);
-}
-
-define_mask_vector! {
-    /// Vector of 16 `mask32` values
-    struct mask32x16([i32 as mask32; 16]);
-}
diff --git a/crates/core_simd/src/vectors_mask64.rs b/crates/core_simd/src/vectors_mask64.rs
deleted file mode 100644
index 554e731ccf2..00000000000
--- a/crates/core_simd/src/vectors_mask64.rs
+++ /dev/null
@@ -1,16 +0,0 @@
-use crate::mask64;
-
-define_mask_vector! {
-    /// Vector of two `mask64` values
-    struct mask64x2([i64 as mask64; 2]);
-}
-
-define_mask_vector! {
-    /// Vector of four `mask64` values
-    struct mask64x4([i64 as mask64; 4]);
-}
-
-define_mask_vector! {
-    /// Vector of eight `mask64` values
-    struct mask64x8([i64 as mask64; 8]);
-}
diff --git a/crates/core_simd/src/vectors_mask8.rs b/crates/core_simd/src/vectors_mask8.rs
deleted file mode 100644
index d038b336104..00000000000
--- a/crates/core_simd/src/vectors_mask8.rs
+++ /dev/null
@@ -1,21 +0,0 @@
-use crate::mask8;
-
-define_mask_vector! {
-    /// Vector of eight `mask8` values
-    struct mask8x8([i8 as mask8; 8]);
-}
-
-define_mask_vector! {
-    /// Vector of 16 `mask8` values
-    struct mask8x16([i8 as mask8; 16]);
-}
-
-define_mask_vector! {
-    /// Vector of 32 `mask8` values
-    struct mask8x32([i8 as mask8; 32]);
-}
-
-define_mask_vector! {
-    /// Vector of 64 `mask8` values
-    struct mask8x64([i8 as mask8; 64]);
-}
diff --git a/crates/core_simd/src/vectors_masksize.rs b/crates/core_simd/src/vectors_masksize.rs
deleted file mode 100644
index a838aee5198..00000000000
--- a/crates/core_simd/src/vectors_masksize.rs
+++ /dev/null
@@ -1,16 +0,0 @@
-use crate::masksize;
-
-define_mask_vector! {
-    /// Vector of two `masksize` values
-    struct masksizex2([isize as masksize; 2]);
-}
-
-define_mask_vector! {
-    /// Vector of four `masksize` values
-    struct masksizex4([isize as masksize; 4]);
-}
-
-define_mask_vector! {
-    /// Vector of eight `masksize` values
-    struct masksizex8([isize as masksize; 8]);
-}
diff --git a/crates/core_simd/src/vectors_u128.rs b/crates/core_simd/src/vectors_u128.rs
index eec7bde1722..522404f133e 100644
--- a/crates/core_simd/src/vectors_u128.rs
+++ b/crates/core_simd/src/vectors_u128.rs
@@ -1,12 +1,16 @@
-define_integer_vector! {
-    /// Vector of two `u128` values
-    struct u128x2([u128; 2]);
-}
-
-define_integer_vector! {
-    /// Vector of four `u128` values
-    struct u128x4([u128; 4]);
-}
+#![allow(non_camel_case_types)]
+
+/// A SIMD vector of containing `LANES` `u128` values.
+#[repr(simd)]
+pub struct SimdU128<const LANES: usize>([u128; LANES]);
+
+impl_integer_vector! { SimdU128, u128 }
+
+/// Vector of two `u128` values
+pub type u128x2 = SimdU128<2>;
+
+/// Vector of four `u128` values
+pub type u128x4 = SimdU128<4>;
 
 from_transmute_x86! { unsafe u128x2 => __m256i }
 //from_transmute_x86! { unsafe u128x4 => __m512i }
diff --git a/crates/core_simd/src/vectors_u16.rs b/crates/core_simd/src/vectors_u16.rs
index 809ab10383c..efe7dea58dc 100644
--- a/crates/core_simd/src/vectors_u16.rs
+++ b/crates/core_simd/src/vectors_u16.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of four `u16` values
-    struct u16x4([u16; 4]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of eight `u16` values
-    struct u16x8([u16; 8]);
-}
+/// A SIMD vector of containing `LANES` `u16` values.
+#[repr(simd)]
+pub struct SimdU16<const LANES: usize>([u16; LANES]);
 
-define_integer_vector! {
-    /// Vector of 16 `u16` values
-    struct u16x16([u16; 16]);
-}
+impl_integer_vector! { SimdU16, u16 }
 
-define_integer_vector! {
-    /// Vector of 32 `u16` values
-    struct u16x32([u16; 32]);
-}
+/// Vector of four `u16` values
+pub type u16x4 = SimdU16<4>;
+
+/// Vector of eight `u16` values
+pub type u16x8 = SimdU16<8>;
+
+/// Vector of 16 `u16` values
+pub type u16x16 = SimdU16<16>;
+
+/// Vector of 32 `u16` values
+pub type u16x32 = SimdU16<32>;
 
 from_transmute_x86! { unsafe u16x8 => __m128i }
 from_transmute_x86! { unsafe u16x16 => __m256i }
diff --git a/crates/core_simd/src/vectors_u32.rs b/crates/core_simd/src/vectors_u32.rs
index b00c63d9058..a6cef5baeb7 100644
--- a/crates/core_simd/src/vectors_u32.rs
+++ b/crates/core_simd/src/vectors_u32.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of two `u32` values
-    struct u32x2([u32; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `u32` values
-    struct u32x4([u32; 4]);
-}
+/// A SIMD vector of containing `LANES` `u32` values.
+#[repr(simd)]
+pub struct SimdU32<const LANES: usize>([u32; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `u32` values
-    struct u32x8([u32; 8]);
-}
+impl_integer_vector! { SimdU32, u32 }
 
-define_integer_vector! {
-    /// Vector of 16 `u32` values
-    struct u32x16([u32; 16]);
-}
+/// Vector of two `u32` values
+pub type u32x2 = SimdU32<2>;
+
+/// Vector of four `u32` values
+pub type u32x4 = SimdU32<4>;
+
+/// Vector of eight `u32` values
+pub type u32x8 = SimdU32<8>;
+
+/// Vector of 16 `u32` values
+pub type u32x16 = SimdU32<16>;
 
 from_transmute_x86! { unsafe u32x4 => __m128i }
 from_transmute_x86! { unsafe u32x8 => __m256i }
diff --git a/crates/core_simd/src/vectors_u64.rs b/crates/core_simd/src/vectors_u64.rs
index 0bcf28ebc26..3982e30f570 100644
--- a/crates/core_simd/src/vectors_u64.rs
+++ b/crates/core_simd/src/vectors_u64.rs
@@ -1,17 +1,19 @@
-define_integer_vector! {
-    /// Vector of two `u64` values
-    struct u64x2([u64; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `u64` values
-    struct u64x4([u64; 4]);
-}
+/// A SIMD vector of containing `LANES` `u64` values.
+#[repr(simd)]
+pub struct SimdU64<const LANES: usize>([u64; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `u64` values
-    struct u64x8([u64; 8]);
-}
+impl_integer_vector! { SimdU64, u64 }
+
+/// Vector of two `u64` values
+pub type u64x2 = SimdU64<2>;
+
+/// Vector of four `u64` values
+pub type u64x4 = SimdU64<4>;
+
+/// Vector of eight `u64` values
+pub type u64x8 = SimdU64<8>;
 
 from_transmute_x86! { unsafe u64x2 => __m128i }
 from_transmute_x86! { unsafe u64x4 => __m256i }
diff --git a/crates/core_simd/src/vectors_u8.rs b/crates/core_simd/src/vectors_u8.rs
index a187bc6f7b4..9cc4eaca47a 100644
--- a/crates/core_simd/src/vectors_u8.rs
+++ b/crates/core_simd/src/vectors_u8.rs
@@ -1,22 +1,22 @@
-define_integer_vector! {
-    /// Vector of eight `u8` values
-    struct u8x8([u8; 8]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of 16 `u8` values
-    struct u8x16([u8; 16]);
-}
+/// A SIMD vector of containing `LANES` `u8` values.
+#[repr(simd)]
+pub struct SimdU8<const LANES: usize>([u8; LANES]);
 
-define_integer_vector! {
-    /// Vector of 32 `u8` values
-    struct u8x32([u8; 32]);
-}
+impl_integer_vector! { SimdU8, u8 }
 
-define_integer_vector! {
-    /// Vector of 64 `u8` values
-    struct u8x64([u8; 64]);
-}
+/// Vector of eight `u8` values
+pub type u8x8 = SimdU8<8>;
+
+/// Vector of 16 `u8` values
+pub type u8x16 = SimdU8<16>;
+
+/// Vector of 32 `u8` values
+pub type u8x32 = SimdU8<32>;
+
+/// Vector of 64 `u8` values
+pub type u8x64 = SimdU8<64>;
 
 from_transmute_x86! { unsafe u8x16 => __m128i }
 from_transmute_x86! { unsafe u8x32 => __m256i }
diff --git a/crates/core_simd/src/vectors_usize.rs b/crates/core_simd/src/vectors_usize.rs
index 84a4b8e509b..c882898f9fb 100644
--- a/crates/core_simd/src/vectors_usize.rs
+++ b/crates/core_simd/src/vectors_usize.rs
@@ -1,17 +1,19 @@
-define_integer_vector! {
-    /// Vector of two `usize` values
-    struct usizex2([usize; 2]);
-}
+#![allow(non_camel_case_types)]
 
-define_integer_vector! {
-    /// Vector of four `usize` values
-    struct usizex4([usize; 4]);
-}
+/// A SIMD vector of containing `LANES` `usize` values.
+#[repr(simd)]
+pub struct SimdUsize<const LANES: usize>([usize; LANES]);
 
-define_integer_vector! {
-    /// Vector of eight `usize` values
-    struct usizex8([usize; 8]);
-}
+impl_integer_vector! { SimdUsize, usize }
+
+/// Vector of two `usize` values
+pub type usizex2 = SimdUsize<2>;
+
+/// Vector of four `usize` values
+pub type usizex4 = SimdUsize<4>;
+
+/// Vector of eight `usize` values
+pub type usizex8 = SimdUsize<8>;
 
 #[cfg(target_pointer_width = "32")]
 from_transmute_x86! { unsafe usizex4 => __m128i }
diff --git a/crates/core_simd/tests/helpers/biteq.rs b/crates/core_simd/tests/helpers/biteq.rs
index f932eba907c..9da2bdfce42 100644
--- a/crates/core_simd/tests/helpers/biteq.rs
+++ b/crates/core_simd/tests/helpers/biteq.rs
@@ -70,7 +70,6 @@ impl_biteq! {
     integer impl BitEq for
         u8, u16, u32, u64, u128, usize,
         i8, i16, i32, i64, i128, isize,
-        core_simd::mask8, core_simd::mask16, core_simd::mask32, core_simd::mask64, core_simd::mask128, core_simd::masksize,
 }
 
 impl_biteq! {
@@ -93,12 +92,6 @@ impl_biteq! {
         core_simd::isizex2, core_simd::isizex4, core_simd::isizex8,
         core_simd::f32x2, core_simd::f32x4, core_simd::f32x8, core_simd::f32x16,
         core_simd::f64x2, core_simd::f64x4, core_simd::f64x8,
-        core_simd::mask8x8,    core_simd::mask8x16,   core_simd::mask8x32,   core_simd::mask8x64,
-        core_simd::mask16x4,   core_simd::mask16x8,   core_simd::mask16x16,  core_simd::mask16x32,
-        core_simd::mask32x2,   core_simd::mask32x4,   core_simd::mask32x8,   core_simd::mask32x16,
-        core_simd::mask64x2,   core_simd::mask64x4,   core_simd::mask64x8,
-        core_simd::mask128x2,  core_simd::mask128x4,
-        core_simd::masksizex2, core_simd::masksizex4, core_simd::masksizex8,
 }
 
 pub(crate) struct BitEqWrapper<'a, T>(pub(crate) &'a T);
diff --git a/crates/core_simd/tests/ops_impl/mask128.rs b/crates/core_simd/tests/ops_impl/mask128.rs
index f0bcdb4d4df..27ba4e2d29f 100644
--- a/crates/core_simd/tests/ops_impl/mask128.rs
+++ b/crates/core_simd/tests/ops_impl/mask128.rs
@@ -1,4 +1,2 @@
-use super::helpers;
-
-mask_tests! { mask128x2, mask128 }
-mask_tests! { mask128x4, mask128 }
+mask_tests! { mask128x2, 2 }
+mask_tests! { mask128x4, 4 }
diff --git a/crates/core_simd/tests/ops_impl/mask16.rs b/crates/core_simd/tests/ops_impl/mask16.rs
index 6f3f8e0ee02..0fe82fa6804 100644
--- a/crates/core_simd/tests/ops_impl/mask16.rs
+++ b/crates/core_simd/tests/ops_impl/mask16.rs
@@ -1,6 +1,4 @@
-use super::helpers;
-
-mask_tests! { mask16x4, mask16 }
-mask_tests! { mask16x8, mask16 }
-mask_tests! { mask16x16, mask16 }
-mask_tests! { mask16x32, mask16 }
+mask_tests! { mask16x4, 4 }
+mask_tests! { mask16x8, 8 }
+mask_tests! { mask16x16, 16 }
+mask_tests! { mask16x32, 32 }
diff --git a/crates/core_simd/tests/ops_impl/mask32.rs b/crates/core_simd/tests/ops_impl/mask32.rs
index 5c35885a2f5..66d987a43ce 100644
--- a/crates/core_simd/tests/ops_impl/mask32.rs
+++ b/crates/core_simd/tests/ops_impl/mask32.rs
@@ -1,6 +1,4 @@
-use super::helpers;
-
-mask_tests! { mask32x2, mask32 }
-mask_tests! { mask32x4, mask32 }
-mask_tests! { mask32x8, mask32 }
-mask_tests! { mask32x16, mask32 }
+mask_tests! { mask32x2, 2 }
+mask_tests! { mask32x4, 4 }
+mask_tests! { mask32x8, 8 }
+mask_tests! { mask32x16, 16 }
diff --git a/crates/core_simd/tests/ops_impl/mask64.rs b/crates/core_simd/tests/ops_impl/mask64.rs
index 88d3211465c..a1f1f67b238 100644
--- a/crates/core_simd/tests/ops_impl/mask64.rs
+++ b/crates/core_simd/tests/ops_impl/mask64.rs
@@ -1,5 +1,3 @@
-use super::helpers;
-
-mask_tests! { mask64x2, mask64 }
-mask_tests! { mask64x4, mask64 }
-mask_tests! { mask64x8, mask64 }
+mask_tests! { mask64x2, 2 }
+mask_tests! { mask64x4, 4 }
+mask_tests! { mask64x8, 8 }
diff --git a/crates/core_simd/tests/ops_impl/mask8.rs b/crates/core_simd/tests/ops_impl/mask8.rs
index fa4bcf09f36..218fa9fe895 100644
--- a/crates/core_simd/tests/ops_impl/mask8.rs
+++ b/crates/core_simd/tests/ops_impl/mask8.rs
@@ -1,6 +1,4 @@
-use super::helpers;
-
-mask_tests! { mask8x8, mask8 }
-mask_tests! { mask8x16, mask8 }
-mask_tests! { mask8x32, mask8 }
-mask_tests! { mask8x64, mask8 }
+mask_tests! { mask8x8, 8 }
+mask_tests! { mask8x16, 16 }
+mask_tests! { mask8x32, 32 }
+mask_tests! { mask8x64, 64 }
diff --git a/crates/core_simd/tests/ops_impl/mask_macros.rs b/crates/core_simd/tests/ops_impl/mask_macros.rs
index e6aee4c1d30..795f9e27c44 100644
--- a/crates/core_simd/tests/ops_impl/mask_macros.rs
+++ b/crates/core_simd/tests/ops_impl/mask_macros.rs
@@ -1,9 +1,9 @@
 macro_rules! mask_tests {
-    { $vector:ident, $scalar:ident } => {
+    { $vector:ident, $lanes:literal } => {
         #[cfg(test)]
         mod $vector {
-            use super::*;
-            use helpers::lanewise::*;
+            use core_simd::$vector as Vector;
+            const LANES: usize = $lanes;
 
             #[cfg(target_arch = "wasm32")]
             use wasm_bindgen_test::*;
@@ -11,15 +11,44 @@ macro_rules! mask_tests {
             #[cfg(target_arch = "wasm32")]
             wasm_bindgen_test_configure!(run_in_browser);
 
-            fn from_slice(slice: &[bool]) -> core_simd::$vector {
-                let mut value = core_simd::$vector::default();
-                let value_slice: &mut [_] = value.as_mut();
-                for (m, b) in value_slice.iter_mut().zip(slice.iter()) {
-                    *m = (*b).into();
+            fn from_slice(slice: &[bool]) -> Vector {
+                let mut value = Vector::default();
+                for (i, b) in slice.iter().take(LANES).enumerate() {
+                    value.set(i, *b);
                 }
                 value
             }
 
+            fn apply_unary_lanewise(x: Vector, f: impl Fn(bool) -> bool) -> Vector {
+                let mut value = Vector::default();
+                for i in 0..LANES {
+                    value.set(i, f(x.test(i)));
+                }
+                value
+            }
+
+            fn apply_binary_lanewise(x: Vector, y: Vector, f: impl Fn(bool, bool) -> bool) -> Vector {
+                let mut value = Vector::default();
+                for i in 0..LANES {
+                    value.set(i, f(x.test(i), y.test(i)));
+                }
+                value
+            }
+
+            fn apply_binary_scalar_lhs_lanewise(x: bool, mut y: Vector, f: impl Fn(bool, bool) -> bool) -> Vector {
+                for i in 0..LANES {
+                    y.set(i, f(x, y.test(i)));
+                }
+                y
+            }
+
+            fn apply_binary_scalar_rhs_lanewise(mut x: Vector, y: bool, f: impl Fn(bool, bool) -> bool) -> Vector {
+                for i in 0..LANES {
+                    x.set(i, f(x.test(i), y));
+                }
+                x
+            }
+
             const A: [bool; 64] = [
                 false, true, false, true, false, false, true, true,
                 false, true, false, true, false, false, true, true,
@@ -41,18 +70,13 @@ macro_rules! mask_tests {
                 false, false, true, true, false, true, false, true,
             ];
 
-            const SET_SCALAR: core_simd::$scalar = core_simd::$scalar::new(true);
-            const UNSET_SCALAR: core_simd::$scalar = core_simd::$scalar::new(false);
-            const SET_VECTOR: core_simd::$vector = core_simd::$vector::splat(SET_SCALAR);
-            const UNSET_VECTOR: core_simd::$vector = core_simd::$vector::splat(UNSET_SCALAR);
-
             #[test]
             #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test)]
             fn bitand() {
                 let a = from_slice(&A);
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitAnd::bitand);
-                assert_biteq!(a & b, expected);
+                assert_eq!(a & b, expected);
             }
 
             #[test]
@@ -62,7 +86,7 @@ macro_rules! mask_tests {
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitAnd::bitand);
                 a &= b;
-                assert_biteq!(a, expected);
+                assert_eq!(a, expected);
             }
 
             #[test]
@@ -70,8 +94,8 @@ macro_rules! mask_tests {
             fn bitand_scalar_rhs() {
                 let a = from_slice(&A);
                 let expected = a;
-                assert_biteq!(a & SET_SCALAR, expected);
-                assert_biteq!(a & UNSET_SCALAR, UNSET_VECTOR);
+                assert_eq!(a & true, expected);
+                assert_eq!(a & false, Vector::splat(false));
             }
 
             #[test]
@@ -79,8 +103,8 @@ macro_rules! mask_tests {
             fn bitand_scalar_lhs() {
                 let a = from_slice(&A);
                 let expected = a;
-                assert_biteq!(SET_SCALAR & a, expected);
-                assert_biteq!(UNSET_SCALAR & a, UNSET_VECTOR);
+                assert_eq!(true & a, expected);
+                assert_eq!(false & a, Vector::splat(false));
             }
 
             #[test]
@@ -88,10 +112,10 @@ macro_rules! mask_tests {
             fn bitand_assign_scalar() {
                 let mut a = from_slice(&A);
                 let expected = a;
-                a &= SET_SCALAR;
-                assert_biteq!(a, expected);
-                a &= UNSET_SCALAR;
-                assert_biteq!(a, UNSET_VECTOR);
+                a &= true;
+                assert_eq!(a, expected);
+                a &= false;
+                assert_eq!(a, Vector::splat(false));
             }
 
             #[test]
@@ -100,7 +124,7 @@ macro_rules! mask_tests {
                 let a = from_slice(&A);
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitOr::bitor);
-                assert_biteq!(a | b, expected);
+                assert_eq!(a | b, expected);
             }
 
             #[test]
@@ -110,23 +134,23 @@ macro_rules! mask_tests {
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitOr::bitor);
                 a |= b;
-                assert_biteq!(a, expected);
+                assert_eq!(a, expected);
             }
 
             #[test]
             #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test)]
             fn bitor_scalar_rhs() {
                 let a = from_slice(&A);
-                assert_biteq!(a | UNSET_SCALAR, a);
-                assert_biteq!(a | SET_SCALAR, SET_VECTOR);
+                assert_eq!(a | false, a);
+                assert_eq!(a | true, Vector::splat(true));
             }
 
             #[test]
             #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test)]
             fn bitor_scalar_lhs() {
                 let a = from_slice(&A);
-                assert_biteq!(UNSET_SCALAR | a, a);
-                assert_biteq!(SET_SCALAR | a, SET_VECTOR);
+                assert_eq!(false | a, a);
+                assert_eq!(true | a, Vector::splat(true));
             }
 
             #[test]
@@ -134,10 +158,10 @@ macro_rules! mask_tests {
             fn bitor_assign_scalar() {
                 let mut a = from_slice(&A);
                 let expected = a;
-                a |= UNSET_SCALAR;
-                assert_biteq!(a, expected);
-                a |= SET_SCALAR;
-                assert_biteq!(a, SET_VECTOR);
+                a |= false;
+                assert_eq!(a, expected);
+                a |= true;
+                assert_eq!(a, Vector::splat(true));
             }
 
             #[test]
@@ -146,7 +170,7 @@ macro_rules! mask_tests {
                 let a = from_slice(&A);
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitXor::bitxor);
-                assert_biteq!(a ^ b, expected);
+                assert_eq!(a ^ b, expected);
             }
 
             #[test]
@@ -156,25 +180,25 @@ macro_rules! mask_tests {
                 let b = from_slice(&B);
                 let expected = apply_binary_lanewise(a, b, core::ops::BitXor::bitxor);
                 a ^= b;
-                assert_biteq!(a, expected);
+                assert_eq!(a, expected);
             }
 
             #[test]
             #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test)]
             fn bitxor_scalar_rhs() {
                 let a = from_slice(&A);
-                let expected = apply_binary_scalar_rhs_lanewise(a, SET_SCALAR, core::ops::BitXor::bitxor);
-                assert_biteq!(a ^ UNSET_SCALAR, a);
-                assert_biteq!(a ^ SET_SCALAR, expected);
+                let expected = apply_binary_scalar_rhs_lanewise(a, true, core::ops::BitXor::bitxor);
+                assert_eq!(a ^ false, a);
+                assert_eq!(a ^ true, expected);
             }
 
             #[test]
             #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test)]
             fn bitxor_scalar_lhs() {
                 let a = from_slice(&A);
-                let expected = apply_binary_scalar_lhs_lanewise(SET_SCALAR, a, core::ops::BitXor::bitxor);
-                assert_biteq!(UNSET_SCALAR ^ a, a);
-                assert_biteq!(SET_SCALAR ^ a, expected);
+                let expected = apply_binary_scalar_lhs_lanewise(true, a, core::ops::BitXor::bitxor);
+                assert_eq!(false ^ a, a);
+                assert_eq!(true ^ a, expected);
             }
 
             #[test]
@@ -182,11 +206,11 @@ macro_rules! mask_tests {
             fn bitxor_assign_scalar() {
                 let mut a = from_slice(&A);
                 let expected_unset = a;
-                let expected_set = apply_binary_scalar_rhs_lanewise(a, SET_SCALAR, core::ops::BitXor::bitxor);
-                a ^= UNSET_SCALAR;
-                assert_biteq!(a, expected_unset);
-                a ^= SET_SCALAR;
-                assert_biteq!(a, expected_set);
+                let expected_set = apply_binary_scalar_rhs_lanewise(a, true, core::ops::BitXor::bitxor);
+                a ^= false;
+                assert_eq!(a, expected_unset);
+                a ^= true;
+                assert_eq!(a, expected_set);
             }
 
             #[test]
@@ -194,7 +218,7 @@ macro_rules! mask_tests {
             fn not() {
                 let v = from_slice(&A);
                 let expected = apply_unary_lanewise(v, core::ops::Not::not);
-                assert_biteq!(!v, expected);
+                assert_eq!(!v, expected);
             }
         }
     }
diff --git a/crates/core_simd/tests/ops_impl/masksize.rs b/crates/core_simd/tests/ops_impl/masksize.rs
index 76e333f3c15..e0a44d870ca 100644
--- a/crates/core_simd/tests/ops_impl/masksize.rs
+++ b/crates/core_simd/tests/ops_impl/masksize.rs
@@ -1,5 +1,3 @@
-use super::helpers;
-
-mask_tests! { masksizex2, masksize }
-mask_tests! { masksizex4, masksize }
-mask_tests! { masksizex8, masksize }
+mask_tests! { masksizex2, 2 }
+mask_tests! { masksizex4, 4 }
+mask_tests! { masksizex8, 8 }