diff --git a/src/implementation/aarch64/neon.rs b/src/implementation/aarch64/neon.rs
index c629c042..5f94a314 100644
--- a/src/implementation/aarch64/neon.rs
+++ b/src/implementation/aarch64/neon.rs
@@ -82,19 +82,7 @@ impl SimdU8Value {
 
     #[inline]
     unsafe fn load_from(ptr: *const u8) -> Self {
-        // WORKAROUND for https://github.com/rust-lang/stdarch/issues/1148
-        // The vld1q_u8 intrinsic is currently broken, it treats it as individual
-        // byte loads so the compiler sometimes decides it is a better to load
-        // individual bytes to "optimize" a subsequent SIMD shuffle
-        //
-        // This code forces a full 128-bit load.
-        let mut dst = core::mem::MaybeUninit::<uint8x16_t>::uninit();
-        core::ptr::copy_nonoverlapping(
-            ptr.cast::<u8>(),
-            dst.as_mut_ptr().cast::<u8>(),
-            core::mem::size_of::<uint8x16_t>(),
-        );
-        Self::from(dst.assume_init())
+        Self::from(vld1q_u8(ptr))
     }
 
     #[inline]