Apply suggestions from code review

Co-authored-by: Antoine Pitrou <pitrou@free.fr>

Author: AntoinePrv

include/xsimd/arch/xsimd_avx2.hpp

@@ -333,14 +333,16 @@ namespace xsimd
             }
         }
 
-        // bitwise_lshift multiple (constant) specific implementations.
+        // bitwise_lshift with constant bitshifts, for 16-bit elements.
         // Missing implementations are dispacthed to the `batch` overload in xsimd_api.
         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<avx2> req) noexcept
         {
             using uint_t = typename std::make_unsigned<T>::type;
 
+            // AVX2 only supports 16-bit shifts with a uniform bitshift value,
+            // otherwise emulate using 32-bit shifts.
             XSIMD_IF_CONSTEXPR(utils::all_equals(shifts))
             {
                 return bitwise_lshift<shifts.get(0), A>(self, req);
@@ -351,6 +353,7 @@ namespace xsimd
                     batch_constant<uint_t, A, static_cast<uint_t>(Vs)...> {}));
         }
 
+        // bitwise_lshift with constant bitshifts, for 8-bit elements.
         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<avx2> req) noexcept