Get rid of per-lane access in Neon implementation

serge-sans-paille · serge-sans-paille · commit 261f82a1337a · 2025-07-10T10:48:15.000+02:00
diff --git a/include/xsimd/arch/xsimd_neon.hpp b/include/xsimd/arch/xsimd_neon.hpp
@@ -717,16 +717,10 @@ namespace xsimd
             return vnegq_s32(rhs);
         }
 
-        template <class A, class T, detail::enable_sized_unsigned_t<T, 8> = 0>
-        XSIMD_INLINE batch<T, A> neg(batch<T, A> const& rhs, requires_arch<neon>) noexcept
-        {
-            return batch<T, A> { -rhs.get(0), -rhs.get(1) };
-        }
-
-        template <class A, class T, detail::enable_sized_signed_t<T, 8> = 0>
+        template <class A, class T, detail::enable_sized_integral_t<T, 8> = 0>
         XSIMD_INLINE batch<T, A> neg(batch<T, A> const& rhs, requires_arch<neon>) noexcept
         {
-            return batch<T, A> { -rhs.get(0), -rhs.get(1) };
+            return 0 - rhs;
         }
 
         template <class A>

Original file line number	Diff line number	Diff line change
`@@ -717,16 +717,10 @@ namespace xsimd`
`717`	`717`	`return vnegq_s32(rhs);`
`718`	`718`	`}`
`719`	`719`
`720`		`- template <class A, class T, detail::enable_sized_unsigned_t<T, 8> = 0>`
`721`		`- XSIMD_INLINE batch<T, A> neg(batch<T, A> const& rhs, requires_arch<neon>) noexcept`
`722`		`- {`
`723`		`- return batch<T, A> { -rhs.get(0), -rhs.get(1) };`
`724`		`- }`
`725`		`-`
`726`		`- template <class A, class T, detail::enable_sized_signed_t<T, 8> = 0>`
	`720`	`+ template <class A, class T, detail::enable_sized_integral_t<T, 8> = 0>`
`727`	`721`	`XSIMD_INLINE batch<T, A> neg(batch<T, A> const& rhs, requires_arch<neon>) noexcept`
`728`	`722`	`{`
`729`		`- return batch<T, A> { -rhs.get(0), -rhs.get(1) };`
	`723`	`+ return 0 - rhs;`
`730`	`724`	`}`
`731`	`725`
`732`	`726`	`template <class A>`