Enable SSE2 fallback for signed integers

Author: AntoinePrv

include/xsimd/arch/xsimd_sse2.hpp

@@ -329,27 +329,32 @@ namespace xsimd
 
         // bitwise_lshift multiple (constant)
         // Missing implementations are dispacthed to the `batch` overload in xsimd_api.
-        template <class A, uint16_t... Vs>
-        XSIMD_INLINE batch<uint16_t, A> bitwise_lshift(
-            batch<uint16_t, A> const& self, batch_constant<uint16_t, A, Vs...> shifts, requires_arch<sse2> req) noexcept
+        template <class T, class A, T... Vs, detail::enable_sized_integral_t<T, 2> = 0>
+        XSIMD_INLINE batch<T, A> bitwise_lshift(
+            batch<T, A> const& self, batch_constant<T, A, Vs...> shifts, requires_arch<sse2> req) noexcept
         {
             XSIMD_IF_CONSTEXPR(utils::all_equals(shifts))
             {
                 return bitwise_lshift<shifts.get(0), A>(self, req);
             }
-            constexpr auto mults = batch_constant<uint16_t, A, static_cast<uint16_t>(1u << Vs)...>();
+            constexpr auto mults = batch_constant<T, A, static_cast<T>(1u << Vs)...>();
             return _mm_mullo_epi16(self, mults.as_batch());
         }
 
-        template <class A, uint8_t... Vs>
-        XSIMD_INLINE batch<uint8_t, A> bitwise_lshift(
-            batch<uint8_t, A> const& self, batch_constant<uint8_t, A, Vs...> shifts, requires_arch<sse2> req) noexcept
+        template <class T, class A, T... Vs, detail::enable_sized_integral_t<T, 1> = 0>
+        XSIMD_INLINE batch<T, A> bitwise_lshift(
+            batch<T, A> const& self, batch_constant<T, A, Vs...> shifts, requires_arch<sse2> req) noexcept
         {
+            using uint_t = typename std::make_unsigned<T>::type;
+
             XSIMD_IF_CONSTEXPR(utils::all_equals(shifts))
             {
                 return bitwise_lshift<shifts.get(0), A>(self, req);
             }
-            return utils::bitwise_lshift_as_twice_larger<uint8_t, uint16_t>(self, shifts);
+            return bitwise_cast<T>(
+                utils::bitwise_lshift_as_twice_larger<uint_t, uint16_t>(
+                    bitwise_cast<uint_t>(self),
+                    batch_constant<uint_t, A, static_cast<uint_t>(Vs)...> {}));
         }
 
         // bitwise_not