From 8705787a801c980515f53a27bba6ca66356e4721 Mon Sep 17 00:00:00 2001
From: Chris Lu <chris@lulabs.net>
Date: Mon, 27 Oct 2025 22:54:30 -0500
Subject: [PATCH 1/2] Fix template argument order for batch_bool loads

`batch_bool<T, A>::load_(un)aligned` dispatches to
`kernel::load_(un)aligned<A>`, which requires that the architecture be
the first template parameter.
---
 include/xsimd/arch/xsimd_avx2.hpp   |  2 +-
 include/xsimd/arch/xsimd_neon.hpp   | 12 ++++++------
 include/xsimd/arch/xsimd_sse4_1.hpp |  2 +-
 3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/include/xsimd/arch/xsimd_avx2.hpp b/include/xsimd/arch/xsimd_avx2.hpp
index 87e5dafae..a9e30b4ab 100644
--- a/include/xsimd/arch/xsimd_avx2.hpp
+++ b/include/xsimd/arch/xsimd_avx2.hpp
@@ -573,7 +573,7 @@ namespace xsimd
 
         // load_unaligned<batch_bool>
 
-        template <class T, class A, class = typename std::enable_if<std::is_integral<T>::value>::type>
+        template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value>::type>
         XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<avx2>) noexcept
         {
             XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
diff --git a/include/xsimd/arch/xsimd_neon.hpp b/include/xsimd/arch/xsimd_neon.hpp
index ad2a25246..5eae38efc 100644
--- a/include/xsimd/arch/xsimd_neon.hpp
+++ b/include/xsimd/arch/xsimd_neon.hpp
@@ -574,39 +574,39 @@ namespace xsimd
         }
 
         /* batch bool version */
-        template <class T, class A, detail::enable_sized_t<T, 1> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 1> = 0>
         XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<neon>) noexcept
         {
             auto vmem = load_unaligned<A>((unsigned char const*)mem, convert<unsigned char> {}, A {});
             return { 0 - vmem.data };
         }
-        template <class T, class A, detail::enable_sized_t<T, 1> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 1> = 0>
         XSIMD_INLINE batch_bool<T, A> load_aligned(bool const* mem, batch_bool<T, A> t, requires_arch<neon> r) noexcept
         {
             return load_unaligned(mem, t, r);
         }
 
-        template <class T, class A, detail::enable_sized_t<T, 2> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 2> = 0>
         XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<neon>) noexcept
         {
             uint16x8_t vmem = vmovl_u8(vld1_u8((unsigned char const*)mem));
             return { 0 - vmem };
         }
 
-        template <class T, class A, detail::enable_sized_t<T, 2> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 2> = 0>
         XSIMD_INLINE batch_bool<T, A> load_aligned(bool const* mem, batch_bool<T, A> t, requires_arch<neon> r) noexcept
         {
             return load_unaligned(mem, t, r);
         }
 
-        template <class T, class A, detail::enable_sized_t<T, 4> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 4> = 0>
         XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<neon>) noexcept
         {
             uint8x8_t tmp = vreinterpret_u8_u32(vset_lane_u32(*(unsigned int*)mem, vdup_n_u32(0), 0));
             return { 0 - vmovl_u16(vget_low_u16(vmovl_u8(tmp))) };
         }
 
-        template <class T, class A, detail::enable_sized_t<T, 4> = 0>
+        template <class A, class T, detail::enable_sized_t<T, 4> = 0>
         XSIMD_INLINE batch_bool<T, A> load_aligned(bool const* mem, batch_bool<T, A> t, requires_arch<neon> r) noexcept
         {
             return load_unaligned(mem, t, r);
diff --git a/include/xsimd/arch/xsimd_sse4_1.hpp b/include/xsimd/arch/xsimd_sse4_1.hpp
index 130d50476..1a64fc878 100644
--- a/include/xsimd/arch/xsimd_sse4_1.hpp
+++ b/include/xsimd/arch/xsimd_sse4_1.hpp
@@ -125,7 +125,7 @@ namespace xsimd
 
         // load_unaligned<batch_bool>
 
-        template <class T, class A, class = typename std::enable_if<(std::is_integral<T>::value && sizeof(T) > 1)>::type>
+        template <class A, class T, class = typename std::enable_if<(std::is_integral<T>::value && sizeof(T) > 1)>::type>
         XSIMD_INLINE batch_bool<T, A> load_unaligned(bool const* mem, batch_bool<T, A>, requires_arch<sse4_1>) noexcept
         {
             // GCC <12 have missing or buggy unaligned load intrinsics; use memcpy to work around this.

From 84a07b4131a7c47f557d408ec76051ed90debdff Mon Sep 17 00:00:00 2001
From: Chris Lu <chris@lulabs.net>
Date: Sun, 2 Nov 2025 23:30:53 -0600
Subject: [PATCH 2/2] remove unneeded overload

---
 include/xsimd/arch/xsimd_neon.hpp | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/include/xsimd/arch/xsimd_neon.hpp b/include/xsimd/arch/xsimd_neon.hpp
index 5eae38efc..27c532d9b 100644
--- a/include/xsimd/arch/xsimd_neon.hpp
+++ b/include/xsimd/arch/xsimd_neon.hpp
@@ -612,13 +612,6 @@ namespace xsimd
             return load_unaligned(mem, t, r);
         }
 
-        template <class A>
-        XSIMD_INLINE batch_bool<float, A> load_aligned(bool const* mem, batch_bool<float, A> t, requires_arch<neon> r) noexcept
-        {
-            uint8x8_t tmp = vreinterpret_u8_u32(vset_lane_u32(*(unsigned int*)mem, vdup_n_u32(0), 0));
-            return { 0 - vmovl_u16(vget_low_u16(vmovl_u8(tmp))) };
-        }
-
         /*********
          * store *
          *********/