Move batch<int8_t, A> swizzle to avx512vbmi with _mm512_permutexvar_epi8

Author: junparser

include/xsimd/arch/xsimd_avx512bw.hpp

@@ -684,18 +684,6 @@ namespace xsimd
             return bitwise_cast<int16_t>(swizzle(bitwise_cast<uint16_t>(self), mask, avx512bw {}));
         }
 
-        template <class A>
-        XSIMD_INLINE batch<uint8_t, A> swizzle(batch<uint8_t, A> const& self, batch<uint8_t, A> mask, requires_arch<avx512bw>) noexcept
-        {
-            return _mm512_shuffle_epi8(self, mask);
-        }
-
-        template <class A>
-        XSIMD_INLINE batch<int8_t, A> swizzle(batch<int8_t, A> const& self, batch<uint8_t, A> mask, requires_arch<avx512bw>) noexcept
-        {
-            return bitwise_cast<int8_t>(swizzle(bitwise_cast<uint8_t>(self), mask, avx512bw {}));
-        }
-
         // swizzle (static version)
         template <class A, uint16_t... Vs>
         XSIMD_INLINE batch<uint16_t, A> swizzle(batch<uint16_t, A> const& self, batch_constant<uint16_t, A, Vs...> mask, requires_arch<avx512bw>) noexcept
@@ -709,18 +697,6 @@ namespace xsimd
             return swizzle(self, mask.as_batch(), avx512bw {});
         }
 
-        template <class A, uint8_t... Vs>
-        XSIMD_INLINE batch<uint8_t, A> swizzle(batch<uint8_t, A> const& self, batch_constant<uint8_t, A, Vs...> mask, requires_arch<avx512bw>) noexcept
-        {
-            return swizzle(self, mask.as_batch(), avx512bw {});
-        }
-
-        template <class A, uint8_t... Vs>
-        XSIMD_INLINE batch<int8_t, A> swizzle(batch<int8_t, A> const& self, batch_constant<uint8_t, A, Vs...> mask, requires_arch<avx512bw>) noexcept
-        {
-            return swizzle(self, mask.as_batch(), avx512bw {});
-        }
-
         // zip_hi
         template <class A, class T, class = typename std::enable_if<std::is_integral<T>::value, void>::type>
         XSIMD_INLINE batch<T, A> zip_hi(batch<T, A> const& self, batch<T, A> const& other, requires_arch<avx512bw>) noexcept

include/xsimd/arch/xsimd_avx512vbmi.hpp

@@ -74,6 +74,31 @@ namespace xsimd
             return _mm512_maskz_permutexvar_epi8(mask, _mm512_load_epi32(slide_pattern.data()), x);
         }
 
+        // swizzle (dynamic version)
+        template <class A>
+        XSIMD_INLINE batch<uint8_t, A> swizzle(batch<uint8_t, A> const& self, batch<uint8_t, A> mask, requires_arch<avx512vbmi>) noexcept
+        {
+            return _mm512_permutexvar_epi8(mask, self);
+        }
+
+        template <class A>
+        XSIMD_INLINE batch<int8_t, A> swizzle(batch<int8_t, A> const& self, batch<uint8_t, A> mask, requires_arch<avx512vbmi>) noexcept
+        {
+            return bitwise_cast<int8_t>(swizzle(bitwise_cast<uint8_t>(self), mask, avx512vbmi {}));
+        }
+
+        // swizzle (static version)
+        template <class A, uint8_t... Vs>
+        XSIMD_INLINE batch<uint8_t, A> swizzle(batch<uint8_t, A> const& self, batch_constant<uint8_t, A, Vs...> mask, requires_arch<avx512vbmi>) noexcept
+        {
+            return swizzle(self, mask.as_batch(), avx512vbmi {});
+        }
+
+        template <class A, uint8_t... Vs>
+        XSIMD_INLINE batch<int8_t, A> swizzle(batch<int8_t, A> const& self, batch_constant<uint8_t, A, Vs...> mask, requires_arch<avx512vbmi>) noexcept
+        {
+            return swizzle(self, mask.as_batch(), avx512vbmi {});
+        }
     }
 }