Provide a generic implementation of swizzle with constant mask

As a side effect, improve test coverage for swizzle to all integer
types.

Author: serge-sans-paille

include/xsimd/arch/common/xsimd_common_details.hpp

@@ -90,6 +90,9 @@ namespace xsimd
     XSIMD_INLINE std::pair<batch<T, A>, batch<T, A>> sincos(batch<T, A> const& self) noexcept;
     template <class T, class A>
     XSIMD_INLINE batch<T, A> sqrt(batch<T, A> const& self) noexcept;
+    template <class T, class A, class Vt, Vt... Values>
+    XSIMD_INLINE typename std::enable_if<std::is_arithmetic<T>::value, batch<T, A>>::type
+    swizzle(batch<T, A> const& x, batch_constant<Vt, A, Values...> mask) noexcept;
     template <class T, class A>
     XSIMD_INLINE batch<T, A> tan(batch<T, A> const& self) noexcept;
     template <class T, class A>

include/xsimd/arch/common/xsimd_common_memory.hpp

@@ -341,7 +341,7 @@ namespace xsimd
                 }
             };
 
-            return swizzle(self, make_batch_constant<as_unsigned_integer_t<T>, rotate_generator, A>(), A {});
+            return swizzle(self, make_batch_constant<as_unsigned_integer_t<T>, rotate_generator, A>());
         }
 
         template <size_t N, class A, class T>
@@ -362,7 +362,7 @@ namespace xsimd
                 }
             };
 
-            return swizzle(self, make_batch_constant<as_unsigned_integer_t<T>, rotate_generator, A>(), A {});
+            return swizzle(self, make_batch_constant<as_unsigned_integer_t<T>, rotate_generator, A>());
         }
 
         template <size_t N, class A, class T>
@@ -611,6 +611,15 @@ namespace xsimd
             return batch<T, A>::load_aligned(out_buffer);
         }
 
+        template <class A, class T, class ITy, ITy... Is>
+        XSIMD_INLINE batch<T, A> swizzle(batch<T, A> const& self, batch_constant<ITy, A, Is...>, requires_arch<common>) noexcept
+        {
+            constexpr size_t size = batch<T, A>::size;
+            alignas(A::alignment()) T self_buffer[size];
+            store_aligned(&self_buffer[0], self);
+            return {self_buffer[Is]...};
+        }
+
         template <class A, class T, class ITy>
         XSIMD_INLINE batch<std::complex<T>, A> swizzle(batch<std::complex<T>, A> const& self, batch<ITy, A> mask, requires_arch<common>) noexcept
         {

test/test_utils.hpp

@@ -591,13 +591,7 @@ namespace xsimd
 #define BATCH_TYPES BATCH_INT_TYPES, BATCH_FLOAT_TYPES
 #define BATCH_MATH_TYPES xsimd::batch<int32_t>, BATCH_FLOAT_TYPES
 
-#if !XSIMD_WITH_AVX || XSIMD_WITH_AVX2
-#define BATCH_SWIZZLE_TAIL , xsimd::batch<uint32_t>, xsimd::batch<int32_t>, xsimd::batch<uint64_t>, xsimd::batch<int64_t>
-#else
-#define BATCH_SWIZZLE_TAIL
-#endif
-
-#define BATCH_SWIZZLE_TYPES BATCH_FLOAT_TYPES, BATCH_COMPLEX_TYPES BATCH_SWIZZLE_TAIL
+#define BATCH_SWIZZLE_TYPES BATCH_FLOAT_TYPES, BATCH_COMPLEX_TYPES, BATCH_INT_TYPES
 
 /********************
  * conversion utils *