diff --git a/crates/core_simd/src/fmt.rs b/crates/core_simd/src/fmt.rs
index 16b8f3b95d9ab..6fa238cfda6ac 100644
--- a/crates/core_simd/src/fmt.rs
+++ b/crates/core_simd/src/fmt.rs
@@ -33,7 +33,10 @@ macro_rules! impl_fmt_trait {
     { $($type:ident => $(($trait:ident, $format:ident)),*;)* } => {
         $( // repeat type
             $( // repeat trait
-                impl<const LANES: usize> core::fmt::$trait for crate::$type<LANES> {
+                impl<const LANES: usize> core::fmt::$trait for crate::$type<LANES>
+                where
+                    Self: crate::LanesAtMost64,
+                {
                     fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
                         $format(self.as_ref(), f)
                     }
diff --git a/crates/core_simd/src/lanes_at_most_64.rs b/crates/core_simd/src/lanes_at_most_64.rs
new file mode 100644
index 0000000000000..63882152b6d9f
--- /dev/null
+++ b/crates/core_simd/src/lanes_at_most_64.rs
@@ -0,0 +1,35 @@
+/// Implemented for bitmask sizes that are supported by the implementation.
+pub trait LanesAtMost64 {}
+
+macro_rules! impl_for {
+    { $name:ident } => {
+        impl LanesAtMost64 for $name<1> {}
+        impl LanesAtMost64 for $name<2> {}
+        impl LanesAtMost64 for $name<4> {}
+        impl LanesAtMost64 for $name<8> {}
+        impl LanesAtMost64 for $name<16> {}
+        impl LanesAtMost64 for $name<32> {}
+        impl LanesAtMost64 for $name<64> {}
+    }
+}
+
+use crate::*;
+
+impl_for! { SimdU8 }
+impl_for! { SimdU16 }
+impl_for! { SimdU32 }
+impl_for! { SimdU64 }
+impl_for! { SimdU128 }
+impl_for! { SimdUsize }
+
+impl_for! { SimdI8 }
+impl_for! { SimdI16 }
+impl_for! { SimdI32 }
+impl_for! { SimdI64 }
+impl_for! { SimdI128 }
+impl_for! { SimdIsize }
+
+impl_for! { SimdF32 }
+impl_for! { SimdF64 }
+
+impl_for! { BitMask }
diff --git a/crates/core_simd/src/lib.rs b/crates/core_simd/src/lib.rs
index de9cb566022e2..e2e4864cc758f 100644
--- a/crates/core_simd/src/lib.rs
+++ b/crates/core_simd/src/lib.rs
@@ -14,6 +14,9 @@ mod intrinsics;
 mod ops;
 mod round;
 
+mod lanes_at_most_64;
+pub use lanes_at_most_64::*;
+
 mod masks;
 pub use masks::*;
 
diff --git a/crates/core_simd/src/macros.rs b/crates/core_simd/src/macros.rs
index 0abafe71f50da..5328f22b42ab6 100644
--- a/crates/core_simd/src/macros.rs
+++ b/crates/core_simd/src/macros.rs
@@ -29,7 +29,7 @@ macro_rules! from_transmute_x86 {
 /// 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> {
+        impl<const LANES: usize> $name<LANES> where Self: crate::LanesAtMost64 {
             /// Construct a SIMD vector by setting all lanes to the given value.
             pub const fn splat(value: $type) -> Self {
                 Self([value; LANES])
@@ -72,23 +72,23 @@ macro_rules! impl_vector {
             }
         }
 
-        impl<const LANES: usize> Copy for $name<LANES> {}
+        impl<const LANES: usize> Copy for $name<LANES> where Self: crate::LanesAtMost64 {}
 
-        impl<const LANES: usize> Clone for $name<LANES> {
+        impl<const LANES: usize> Clone for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn clone(&self) -> Self {
                 *self
             }
         }
 
-        impl<const LANES: usize> Default for $name<LANES> {
+        impl<const LANES: usize> Default for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn default() -> Self {
                 Self::splat(<$type>::default())
             }
         }
 
-        impl<const LANES: usize> PartialEq for $name<LANES> {
+        impl<const LANES: usize> PartialEq for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn eq(&self, other: &Self) -> bool {
                 // TODO use SIMD equality
@@ -96,7 +96,7 @@ macro_rules! impl_vector {
             }
         }
 
-        impl<const LANES: usize> PartialOrd for $name<LANES> {
+        impl<const LANES: usize> PartialOrd for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
                 // TODO use SIMD equalitya
@@ -105,14 +105,14 @@ macro_rules! impl_vector {
         }
 
         // array references
-        impl<const LANES: usize> AsRef<[$type; LANES]> for $name<LANES> {
+        impl<const LANES: usize> AsRef<[$type; LANES]> for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn as_ref(&self) -> &[$type; LANES] {
                 &self.0
             }
         }
 
-        impl<const LANES: usize> AsMut<[$type; LANES]> for $name<LANES> {
+        impl<const LANES: usize> AsMut<[$type; LANES]> for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn as_mut(&mut self) -> &mut [$type; LANES] {
                 &mut self.0
@@ -120,14 +120,14 @@ macro_rules! impl_vector {
         }
 
         // slice references
-        impl<const LANES: usize> AsRef<[$type]> for $name<LANES> {
+        impl<const LANES: usize> AsRef<[$type]> for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn as_ref(&self) -> &[$type] {
                 &self.0
             }
         }
 
-        impl<const LANES: usize> AsMut<[$type]> for $name<LANES> {
+        impl<const LANES: usize> AsMut<[$type]> for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn as_mut(&mut self) -> &mut [$type] {
                 &mut self.0
@@ -135,14 +135,14 @@ macro_rules! impl_vector {
         }
 
         // vector/array conversion
-        impl<const LANES: usize> From<[$type; LANES]> for $name<LANES> {
+        impl<const LANES: usize> From<[$type; LANES]> for $name<LANES> where Self: crate::LanesAtMost64 {
             fn from(array: [$type; LANES]) -> Self {
                 Self(array)
             }
         }
 
         // splat
-        impl<const LANES: usize> From<$type> for $name<LANES> {
+        impl<const LANES: usize> From<$type> for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn from(value: $type) -> Self {
                 Self::splat(value)
@@ -158,9 +158,9 @@ macro_rules! impl_integer_vector {
     { $name:ident, $type:ty } => {
         impl_vector! { $name, $type }
 
-        impl<const LANES: usize> Eq for $name<LANES> {}
+        impl<const LANES: usize> Eq for $name<LANES> where Self: crate::LanesAtMost64 {}
 
-        impl<const LANES: usize> Ord for $name<LANES> {
+        impl<const LANES: usize> Ord for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn cmp(&self, other: &Self) -> core::cmp::Ordering {
                 // TODO use SIMD cmp
@@ -168,7 +168,7 @@ macro_rules! impl_integer_vector {
             }
         }
 
-        impl<const LANES: usize> core::hash::Hash for $name<LANES> {
+        impl<const LANES: usize> core::hash::Hash for $name<LANES> where Self: crate::LanesAtMost64 {
             #[inline]
             fn hash<H>(&self, state: &mut H)
             where
@@ -187,7 +187,11 @@ macro_rules! impl_float_vector {
     { $name:ident, $type:ty, $bits_ty:ident } => {
         impl_vector! { $name, $type }
 
-        impl<const LANES: usize> $name<LANES> {
+        impl<const LANES: usize> $name<LANES>
+        where
+            Self: crate::LanesAtMost64,
+            crate::$bits_ty<LANES>: crate::LanesAtMost64,
+        {
             /// Raw transmutation to an unsigned integer vector type with the
             /// same size and number of lanes.
             #[inline]
diff --git a/crates/core_simd/src/masks/bitmask.rs b/crates/core_simd/src/masks/bitmask.rs
index 51ed8037043da..b9b1160a3f38c 100644
--- a/crates/core_simd/src/masks/bitmask.rs
+++ b/crates/core_simd/src/masks/bitmask.rs
@@ -1,12 +1,4 @@
-/// Implemented for bitmask sizes that are supported by the implementation.
-pub trait LanesAtMost64 {}
-impl LanesAtMost64 for BitMask<1> {}
-impl LanesAtMost64 for BitMask<2> {}
-impl LanesAtMost64 for BitMask<4> {}
-impl LanesAtMost64 for BitMask<8> {}
-impl LanesAtMost64 for BitMask<16> {}
-impl LanesAtMost64 for BitMask<32> {}
-impl LanesAtMost64 for BitMask<64> {}
+use crate::LanesAtMost64;
 
 /// A mask where each lane is represented by a single bit.
 #[derive(Copy, Clone, Debug)]
diff --git a/crates/core_simd/src/masks/full_masks.rs b/crates/core_simd/src/masks/full_masks.rs
index d7c4af47727ca..fa93d252df464 100644
--- a/crates/core_simd/src/masks/full_masks.rs
+++ b/crates/core_simd/src/masks/full_masks.rs
@@ -16,11 +16,31 @@ impl core::fmt::Display for TryFromMaskError {
 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)]
+        #[derive(Default, PartialEq, PartialOrd, Eq, Ord, Hash)]
         #[repr(transparent)]
-        pub struct $name<const $lanes: usize>($type);
+        pub struct $name<const $lanes: usize>($type)
+        where
+            $type: crate::LanesAtMost64;
+
+        impl<const LANES: usize> Copy for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {}
 
-        impl<const $lanes: usize> $name<$lanes> {
+        impl<const LANES: usize> Clone for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
+            #[inline]
+            fn clone(&self) -> Self {
+                *self
+            }
+        }
+
+        impl<const $lanes: usize> $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             /// Construct a mask by setting all lanes to the given value.
             pub fn splat(value: bool) -> Self {
                 Self(<$type>::splat(
@@ -57,13 +77,19 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const $lanes: usize> core::convert::From<bool> for $name<$lanes> {
+        impl<const $lanes: usize> core::convert::From<bool> for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             fn from(value: bool) -> Self {
                 Self::splat(value)
             }
         }
 
-        impl<const $lanes: usize> core::convert::TryFrom<$type> for $name<$lanes> {
+        impl<const $lanes: usize> core::convert::TryFrom<$type> for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Error = TryFromMaskError;
             fn try_from(value: $type) -> Result<Self, Self::Error> {
                 if value.as_slice().iter().all(|x| *x == 0 || *x == -1) {
@@ -74,7 +100,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const $lanes: usize> core::convert::From<$name<$lanes>> for $type {
+        impl<const $lanes: usize> core::convert::From<$name<$lanes>> for $type
+        where
+            $type: crate::LanesAtMost64,
+        {
             fn from(value: $name<$lanes>) -> Self {
                 value.0
             }
@@ -82,6 +111,7 @@ macro_rules! define_mask {
 
         impl<const $lanes: usize> core::convert::From<crate::BitMask<$lanes>> for $name<$lanes>
         where
+            $type: crate::LanesAtMost64,
             crate::BitMask<$lanes>: crate::LanesAtMost64,
         {
             fn from(value: crate::BitMask<$lanes>) -> Self {
@@ -96,6 +126,7 @@ macro_rules! define_mask {
 
         impl<const $lanes: usize> core::convert::From<$name<$lanes>> for crate::BitMask<$lanes>
         where
+            $type: crate::LanesAtMost64,
             crate::BitMask<$lanes>: crate::LanesAtMost64,
         {
             fn from(value: $name<$lanes>) -> Self {
@@ -108,7 +139,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const $lanes: usize> core::fmt::Debug for $name<$lanes> {
+        impl<const $lanes: usize> core::fmt::Debug for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
                 f.debug_list()
                     .entries((0..LANES).map(|lane| self.test(lane)))
@@ -116,31 +150,46 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const $lanes: usize> core::fmt::Binary for $name<$lanes> {
+        impl<const $lanes: usize> core::fmt::Binary for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             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> {
+        impl<const $lanes: usize> core::fmt::Octal for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             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> {
+        impl<const $lanes: usize> core::fmt::LowerHex for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             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> {
+        impl<const $lanes: usize> core::fmt::UpperHex for $name<$lanes>
+        where
+            $type: crate::LanesAtMost64,
+        {
             fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
                 core::fmt::UpperHex::fmt(&self.0, f)
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitAnd for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitand(self, rhs: Self) -> Self {
@@ -148,7 +197,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitAnd<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitand(self, rhs: bool) -> Self {
@@ -156,7 +208,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd<$name<LANES>> for bool {
+        impl<const LANES: usize> core::ops::BitAnd<$name<LANES>> for bool
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitand(self, rhs: $name<LANES>) -> $name<LANES> {
@@ -164,7 +219,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitOr for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitor(self, rhs: Self) -> Self {
@@ -172,7 +230,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitOr<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitor(self, rhs: bool) -> Self {
@@ -180,7 +241,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr<$name<LANES>> for bool {
+        impl<const LANES: usize> core::ops::BitOr<$name<LANES>> for bool
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitor(self, rhs: $name<LANES>) -> $name<LANES> {
@@ -188,7 +252,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitXor for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitxor(self, rhs: Self) -> Self::Output {
@@ -196,7 +263,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitXor<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitxor(self, rhs: bool) -> Self::Output {
@@ -204,7 +274,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor<$name<LANES>> for bool {
+        impl<const LANES: usize> core::ops::BitXor<$name<LANES>> for bool
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitxor(self, rhs: $name<LANES>) -> Self::Output {
@@ -212,7 +285,10 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::Not for $name<LANES> {
+        impl<const LANES: usize> core::ops::Not for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn not(self) -> Self::Output {
@@ -220,42 +296,60 @@ macro_rules! define_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAndAssign for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitAndAssign for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitand_assign(&mut self, rhs: Self) {
                 self.0 &= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAndAssign<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitAndAssign<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitand_assign(&mut self, rhs: bool) {
                 *self &= Self::splat(rhs);
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOrAssign for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitOrAssign for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitor_assign(&mut self, rhs: Self) {
                 self.0 |= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOrAssign<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitOrAssign<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitor_assign(&mut self, rhs: bool) {
                 *self |= Self::splat(rhs);
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXorAssign for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitXorAssign for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitxor_assign(&mut self, rhs: Self) {
                 self.0 ^= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXorAssign<bool> for $name<LANES> {
+        impl<const LANES: usize> core::ops::BitXorAssign<bool> for $name<LANES>
+        where
+            $type: crate::LanesAtMost64,
+        {
             #[inline]
             fn bitxor_assign(&mut self, rhs: bool) {
                 *self ^= Self::splat(rhs);
@@ -291,11 +385,11 @@ define_mask! {
 define_mask! {
     /// A mask equivalent to [SimdI128](crate::SimdI128), where all bits in the lane must be either set
     /// or unset.
-    struct SimdMask128<const LANES: usize>(crate::SimdI64<LANES>);
+    struct SimdMask128<const LANES: usize>(crate::SimdI128<LANES>);
 }
 
 define_mask! {
     /// A mask equivalent to [SimdIsize](crate::SimdIsize), where all bits in the lane must be either set
     /// or unset.
-    struct SimdMaskSize<const LANES: usize>(crate::SimdI64<LANES>);
+    struct SimdMaskSize<const LANES: usize>(crate::SimdIsize<LANES>);
 }
diff --git a/crates/core_simd/src/masks/mod.rs b/crates/core_simd/src/masks/mod.rs
index e51052f53f2f5..7d7f7af627df2 100644
--- a/crates/core_simd/src/masks/mod.rs
+++ b/crates/core_simd/src/masks/mod.rs
@@ -7,16 +7,22 @@ pub use full_masks::*;
 mod bitmask;
 pub use bitmask::*;
 
+use crate::LanesAtMost64;
+
 macro_rules! define_opaque_mask {
     {
         $(#[$attr:meta])*
         struct $name:ident<const $lanes:ident: usize>($inner_ty:ty);
+        @bits $bits_ty:ty
     } => {
         $(#[$attr])*
         #[allow(non_camel_case_types)]
-        pub struct $name<const $lanes: usize>($inner_ty) where BitMask<LANES>: LanesAtMost64;
+        pub struct $name<const $lanes: usize>($inner_ty) where $bits_ty: LanesAtMost64;
 
-        impl<const $lanes: usize> $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64
+        {
             /// Construct a mask by setting all lanes to the given value.
             pub fn splat(value: bool) -> Self {
                 Self(<$inner_ty>::splat(value))
@@ -43,6 +49,7 @@ macro_rules! define_opaque_mask {
 
         impl<const $lanes: usize> From<BitMask<$lanes>> for $name<$lanes>
         where
+            $bits_ty: LanesAtMost64,
             BitMask<$lanes>: LanesAtMost64,
         {
             fn from(value: BitMask<$lanes>) -> Self {
@@ -52,6 +59,7 @@ macro_rules! define_opaque_mask {
 
         impl<const $lanes: usize> From<$name<$lanes>> for crate::BitMask<$lanes>
         where
+            $bits_ty: LanesAtMost64,
             BitMask<$lanes>: LanesAtMost64,
         {
             fn from(value: $name<$lanes>) -> Self {
@@ -61,7 +69,7 @@ macro_rules! define_opaque_mask {
 
         impl<const $lanes: usize> From<$inner_ty> for $name<$lanes>
         where
-            BitMask<$lanes>: LanesAtMost64,
+            $bits_ty: LanesAtMost64,
         {
             fn from(value: $inner_ty) -> Self {
                 Self(value)
@@ -70,50 +78,72 @@ macro_rules! define_opaque_mask {
 
         impl<const $lanes: usize> From<$name<$lanes>> for $inner_ty
         where
-            BitMask<$lanes>: LanesAtMost64,
+            $bits_ty: LanesAtMost64,
         {
             fn from(value: $name<$lanes>) -> Self {
                 value.0
             }
         }
 
-        impl<const $lanes: usize> Copy for $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {}
+        impl<const $lanes: usize> Copy for $name<$lanes>
+        where
+            $inner_ty: Copy,
+            $bits_ty: LanesAtMost64,
+        {}
 
-        impl<const $lanes: usize> Clone for $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> Clone for $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn clone(&self) -> Self {
                 *self
             }
         }
 
-        impl<const $lanes: usize> Default for $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> Default for $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn default() -> Self {
                 Self::splat(false)
             }
         }
 
-        impl<const $lanes: usize> PartialEq for $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> PartialEq for $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn eq(&self, other: &Self) -> bool {
                 self.0 == other.0
             }
         }
 
-        impl<const $lanes: usize> PartialOrd for $name<$lanes> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> PartialOrd for $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[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> where BitMask<$lanes>: LanesAtMost64 {
+        impl<const $lanes: usize> core::fmt::Debug for $name<$lanes>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
                 core::fmt::Debug::fmt(&self.0, f)
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitAnd for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitand(self, rhs: Self) -> Self {
@@ -121,7 +151,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitAnd<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitand(self, rhs: bool) -> Self {
@@ -129,7 +162,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAnd<$name<LANES>> for bool where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitAnd<$name<LANES>> for bool
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitand(self, rhs: $name<LANES>) -> $name<LANES> {
@@ -137,7 +173,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitOr for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitor(self, rhs: Self) -> Self {
@@ -145,7 +184,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitOr<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitor(self, rhs: bool) -> Self {
@@ -153,7 +195,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOr<$name<LANES>> for bool where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitOr<$name<LANES>> for bool
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitor(self, rhs: $name<LANES>) -> $name<LANES> {
@@ -161,7 +206,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitXor for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitxor(self, rhs: Self) -> Self::Output {
@@ -169,7 +217,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitXor<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = Self;
             #[inline]
             fn bitxor(self, rhs: bool) -> Self::Output {
@@ -177,7 +228,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXor<$name<LANES>> for bool where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitXor<$name<LANES>> for bool
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn bitxor(self, rhs: $name<LANES>) -> Self::Output {
@@ -185,7 +239,10 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::Not for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::Not for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             type Output = $name<LANES>;
             #[inline]
             fn not(self) -> Self::Output {
@@ -193,42 +250,60 @@ macro_rules! define_opaque_mask {
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAndAssign for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitAndAssign for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitand_assign(&mut self, rhs: Self) {
                 self.0 &= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitAndAssign<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitAndAssign<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitand_assign(&mut self, rhs: bool) {
                 *self &= Self::splat(rhs);
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOrAssign for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitOrAssign for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitor_assign(&mut self, rhs: Self) {
                 self.0 |= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitOrAssign<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitOrAssign<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitor_assign(&mut self, rhs: bool) {
                 *self |= Self::splat(rhs);
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXorAssign for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitXorAssign for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitxor_assign(&mut self, rhs: Self) {
                 self.0 ^= rhs.0;
             }
         }
 
-        impl<const LANES: usize> core::ops::BitXorAssign<bool> for $name<LANES> where BitMask<LANES>: LanesAtMost64 {
+        impl<const LANES: usize> core::ops::BitXorAssign<bool> for $name<LANES>
+        where
+            $bits_ty: LanesAtMost64,
+        {
             #[inline]
             fn bitxor_assign(&mut self, rhs: bool) {
                 *self ^= Self::splat(rhs);
@@ -242,6 +317,7 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct Mask8<const LANES: usize>(SimdMask8<LANES>);
+    @bits crate::SimdI8<LANES>
 }
 
 define_opaque_mask! {
@@ -249,6 +325,7 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct Mask16<const LANES: usize>(SimdMask16<LANES>);
+    @bits crate::SimdI16<LANES>
 }
 
 define_opaque_mask! {
@@ -256,6 +333,7 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct Mask32<const LANES: usize>(SimdMask32<LANES>);
+    @bits crate::SimdI32<LANES>
 }
 
 define_opaque_mask! {
@@ -263,6 +341,7 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct Mask64<const LANES: usize>(SimdMask64<LANES>);
+    @bits crate::SimdI64<LANES>
 }
 
 define_opaque_mask! {
@@ -270,6 +349,7 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct Mask128<const LANES: usize>(SimdMask128<LANES>);
+    @bits crate::SimdI128<LANES>
 }
 
 define_opaque_mask! {
@@ -277,12 +357,17 @@ define_opaque_mask! {
     ///
     /// The layout of this type is unspecified.
     struct MaskSize<const LANES: usize>(SimdMaskSize<LANES>);
+    @bits crate::SimdIsize<LANES>
 }
 
 macro_rules! implement_mask_ops {
-    { $($vector:ident => $mask:ident,)* } => {
+    { $($vector:ident => $mask:ident ($inner_ty:ident),)* } => {
         $(
-            impl<const LANES: usize> crate::$vector<LANES> where BitMask<LANES>: LanesAtMost64 {
+            impl<const LANES: usize> crate::$vector<LANES>
+            where
+                crate::$vector<LANES>: LanesAtMost64,
+                crate::$inner_ty<LANES>: LanesAtMost64,
+            {
                 /// Test if each lane is equal to the corresponding lane in `other`.
                 #[inline]
                 pub fn lanes_eq(&self, other: &Self) -> $mask<LANES> {
@@ -324,22 +409,22 @@ macro_rules! implement_mask_ops {
 }
 
 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,
+    SimdI8 => Mask8 (SimdI8),
+    SimdI16 => Mask16 (SimdI16),
+    SimdI32 => Mask32 (SimdI32),
+    SimdI64 => Mask64 (SimdI64),
+    SimdI128 => Mask128 (SimdI128),
+    SimdIsize => MaskSize (SimdIsize),
+
+    SimdU8 => Mask8 (SimdI8),
+    SimdU16 => Mask16 (SimdI16),
+    SimdU32 => Mask32 (SimdI32),
+    SimdU64 => Mask64 (SimdI64),
+    SimdU128 => Mask128 (SimdI128),
+    SimdUsize => MaskSize (SimdIsize),
+
+    SimdF32 => Mask32 (SimdI32),
+    SimdF64 => Mask64 (SimdI64),
 }
 
 /// Vector of eight 8-bit masks
diff --git a/crates/core_simd/src/ops.rs b/crates/core_simd/src/ops.rs
index ae7c86c81c50a..1d9e1eeb92cec 100644
--- a/crates/core_simd/src/ops.rs
+++ b/crates/core_simd/src/ops.rs
@@ -1,3 +1,5 @@
+use crate::LanesAtMost64;
+
 /// Checks if the right-hand side argument of a left- or right-shift would cause overflow.
 fn invalid_shift_rhs<T>(rhs: T) -> bool
 where
@@ -12,21 +14,30 @@ where
 macro_rules! impl_ref_ops {
     // binary op
     {
-        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty
+        where
+            $($bound:path: LanesAtMost64,)*
+        {
             type Output = $output:ty;
 
             $(#[$attrs:meta])*
             fn $fn:ident($self_tok:ident, $rhs_arg:ident: $rhs_arg_ty:ty) -> Self::Output $body:tt
         }
     } => {
-        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = $output;
 
             $(#[$attrs])*
             fn $fn($self_tok, $rhs_arg: $rhs_arg_ty) -> Self::Output $body
         }
 
-        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -35,7 +46,10 @@ macro_rules! impl_ref_ops {
             }
         }
 
-        impl<const $lanes: usize> core::ops::$trait<$rhs> for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for &'_ $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -44,7 +58,10 @@ macro_rules! impl_ref_ops {
             }
         }
 
-        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for &'_ $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = <$type as core::ops::$trait<$rhs>>::Output;
 
             $(#[$attrs])*
@@ -56,17 +73,26 @@ macro_rules! impl_ref_ops {
 
     // binary assignment op
     {
-        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident<$rhs:ty> for $type:ty
+        where
+            $($bound:path: LanesAtMost64,)*
+        {
             $(#[$attrs:meta])*
             fn $fn:ident(&mut $self_tok:ident, $rhs_arg:ident: $rhs_arg_ty:ty) $body:tt
         }
     } => {
-        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<$rhs> for $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             $(#[$attrs])*
             fn $fn(&mut $self_tok, $rhs_arg: $rhs_arg_ty) $body
         }
 
-        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type {
+        impl<const $lanes: usize> core::ops::$trait<&'_ $rhs> for $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             $(#[$attrs])*
             fn $fn(&mut $self_tok, $rhs_arg: &$rhs_arg_ty) {
                 core::ops::$trait::$fn($self_tok, *$rhs_arg)
@@ -76,17 +102,26 @@ macro_rules! impl_ref_ops {
 
     // unary op
     {
-        impl<const $lanes:ident: usize> core::ops::$trait:ident for $type:ty {
+        impl<const $lanes:ident: usize> core::ops::$trait:ident for $type:ty
+        where
+            $($bound:path: LanesAtMost64,)*
+        {
             type Output = $output:ty;
             fn $fn:ident($self_tok:ident) -> Self::Output $body:tt
         }
     } => {
-        impl<const $lanes: usize> core::ops::$trait for $type {
+        impl<const $lanes: usize> core::ops::$trait for $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = $output;
             fn $fn($self_tok) -> Self::Output $body
         }
 
-        impl<const $lanes: usize> core::ops::$trait for &'_ $type {
+        impl<const $lanes: usize> core::ops::$trait for &'_ $type
+        where
+            $($bound: LanesAtMost64,)*
+        {
             type Output = <$type as core::ops::$trait>::Output;
             fn $fn($self_tok) -> Self::Output {
                 core::ops::$trait::$fn(*$self_tok)
@@ -130,7 +165,10 @@ macro_rules! impl_op {
 
     { impl Not for $type:ident, $scalar:ty } => {
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::Not for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::Not for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 type Output = Self;
                 fn not(self) -> Self::Output {
                     self ^ Self::splat(!<$scalar>::default())
@@ -141,7 +179,10 @@ macro_rules! impl_op {
 
     { impl Neg for $type:ident, $scalar:ty } => {
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 type Output = Self;
                 fn neg(self) -> Self::Output {
                     Self::splat(0) - self
@@ -152,7 +193,12 @@ macro_rules! impl_op {
 
     { impl Neg for $type:ident, $scalar:ty, @float } => {
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::Neg for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+                crate::SimdU32<LANES>: LanesAtMost64,
+                crate::SimdU64<LANES>: LanesAtMost64,
+            {
                 type Output = Self;
                 fn neg(self) -> Self::Output {
                     // FIXME: Replace this with fneg intrinsic once available.
@@ -166,6 +212,7 @@ macro_rules! impl_op {
     { impl Index for $type:ident, $scalar:ty } => {
         impl<I, const LANES: usize> core::ops::Index<I> for crate::$type<LANES>
         where
+            Self: LanesAtMost64,
             I: core::slice::SliceIndex<[$scalar]>,
         {
             type Output = I::Output;
@@ -177,6 +224,7 @@ macro_rules! impl_op {
 
         impl<I, const LANES: usize> core::ops::IndexMut<I> for crate::$type<LANES>
         where
+            Self: LanesAtMost64,
             I: core::slice::SliceIndex<[$scalar]>,
         {
             fn index_mut(&mut self, index: I) -> &mut Self::Output {
@@ -189,7 +237,10 @@ macro_rules! impl_op {
     // generic binary op with assignment when output is `Self`
     { @binary $type:ident, $scalar:ty, $trait:ident :: $trait_fn:ident, $assign_trait:ident :: $assign_trait_fn:ident, $intrinsic:ident } => {
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::$trait<Self> for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::$trait<Self> for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 type Output = Self;
 
                 #[inline]
@@ -202,7 +253,10 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::$trait<$scalar> for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::$trait<$scalar> for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 type Output = Self;
 
                 #[inline]
@@ -213,7 +267,10 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::$trait<crate::$type<LANES>> for $scalar {
+            impl<const LANES: usize> core::ops::$trait<crate::$type<LANES>> for $scalar
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 type Output = crate::$type<LANES>;
 
                 #[inline]
@@ -224,7 +281,10 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::$assign_trait<Self> for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::$assign_trait<Self> for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 #[inline]
                 fn $assign_trait_fn(&mut self, rhs: Self) {
                     unsafe {
@@ -235,7 +295,10 @@ macro_rules! impl_op {
         }
 
         impl_ref_ops! {
-            impl<const LANES: usize> core::ops::$assign_trait<$scalar> for crate::$type<LANES> {
+            impl<const LANES: usize> core::ops::$assign_trait<$scalar> for crate::$type<LANES>
+            where
+                crate::$type<LANES>: LanesAtMost64,
+            {
                 #[inline]
                 fn $assign_trait_fn(&mut self, rhs: $scalar) {
                     core::ops::$assign_trait::$assign_trait_fn(self, Self::splat(rhs));
@@ -278,7 +341,10 @@ macro_rules! impl_unsigned_int_ops {
 
                 // Integers panic on divide by 0
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Div<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Div<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -303,7 +369,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Div<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Div<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -323,7 +392,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Div<crate::$vector<LANES>> for $scalar {
+                    impl<const LANES: usize> core::ops::Div<crate::$vector<LANES>> for $scalar
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = crate::$vector<LANES>;
 
                         #[inline]
@@ -334,7 +406,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::DivAssign<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::DivAssign<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn div_assign(&mut self, rhs: Self) {
                             *self = *self / rhs;
@@ -343,7 +418,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::DivAssign<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::DivAssign<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn div_assign(&mut self, rhs: $scalar) {
                             *self = *self / rhs;
@@ -353,7 +431,10 @@ macro_rules! impl_unsigned_int_ops {
 
                 // remainder panics on zero divisor
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Rem<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Rem<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -378,7 +459,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Rem<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Rem<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -398,7 +482,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Rem<crate::$vector<LANES>> for $scalar {
+                    impl<const LANES: usize> core::ops::Rem<crate::$vector<LANES>> for $scalar
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = crate::$vector<LANES>;
 
                         #[inline]
@@ -409,7 +496,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::RemAssign<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::RemAssign<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn rem_assign(&mut self, rhs: Self) {
                             *self = *self % rhs;
@@ -418,7 +508,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::RemAssign<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::RemAssign<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn rem_assign(&mut self, rhs: $scalar) {
                             *self = *self % rhs;
@@ -428,7 +521,10 @@ macro_rules! impl_unsigned_int_ops {
 
                 // shifts panic on overflow
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Shl<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Shl<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -447,7 +543,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Shl<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Shl<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -463,7 +562,10 @@ macro_rules! impl_unsigned_int_ops {
 
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::ShlAssign<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::ShlAssign<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn shl_assign(&mut self, rhs: Self) {
                             *self = *self << rhs;
@@ -472,7 +574,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::ShlAssign<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::ShlAssign<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn shl_assign(&mut self, rhs: $scalar) {
                             *self = *self << rhs;
@@ -481,7 +586,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Shr<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Shr<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -500,7 +608,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::Shr<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::Shr<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         type Output = Self;
 
                         #[inline]
@@ -516,7 +627,10 @@ macro_rules! impl_unsigned_int_ops {
 
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::ShrAssign<Self> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::ShrAssign<Self> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn shr_assign(&mut self, rhs: Self) {
                             *self = *self >> rhs;
@@ -525,7 +639,10 @@ macro_rules! impl_unsigned_int_ops {
                 }
 
                 impl_ref_ops! {
-                    impl<const LANES: usize> core::ops::ShrAssign<$scalar> for crate::$vector<LANES> {
+                    impl<const LANES: usize> core::ops::ShrAssign<$scalar> for crate::$vector<LANES>
+                    where
+                        crate::$vector<LANES>: LanesAtMost64,
+                    {
                         #[inline]
                         fn shr_assign(&mut self, rhs: $scalar) {
                             *self = *self >> rhs;
diff --git a/crates/core_simd/src/vectors_u8.rs b/crates/core_simd/src/vectors_u8.rs
index 9cc4eaca47ad4..6cf623f680136 100644
--- a/crates/core_simd/src/vectors_u8.rs
+++ b/crates/core_simd/src/vectors_u8.rs
@@ -2,7 +2,7 @@
 
 /// A SIMD vector of containing `LANES` `u8` values.
 #[repr(simd)]
-pub struct SimdU8<const LANES: usize>([u8; LANES]);
+pub struct SimdU8<const LANES: usize>([u8; LANES]) where Self: crate::LanesAtMost64;
 
 impl_integer_vector! { SimdU8, u8 }