Improve generic version of complex masked store

Use the usual kernel mechanism which allows for specialization.
Leverage existing masked store mechanism instead of implementing a new
one.

Follow-up to #1391

Author: serge-sans-paille

include/xsimd/arch/common/xsimd_common_memory.hpp

@@ -865,6 +865,39 @@ namespace xsimd
             store_complex_aligned<A>(dst, src, A {});
         }
 
+        template <class A, class T, class Mode>
+        XSIMD_INLINE void
+        store_complex_masked(std::complex<T>* mem, batch<std::complex<T>, A> const& src, batch_bool<T, A> mask, Mode mode, requires_arch<common>) noexcept
+        {
+            // Generic fallback: mask and real /imag part are zipped before
+            // calling the generic masked store routine.
+            using mask_register_type = typename batch_bool<T, A>::register_type;
+            mask_register_type nmask = mask.to_native();
+            batch_bool<T, A> lo_mask;
+            batch_bool<T, A> hi_mask;
+
+            // Generic zip_lo/hi of batch_bool depending on native register type {
+            if constexpr (sizeof(mask_register_type) < sizeof(mask))
+            {
+                constexpr mask_register_type lo_bitmask = xsimd::utils::make_low_mask<mask_register_type>(src.size / 2);
+                constexpr mask_register_type hi_bitmask = lo_bitmask << (src.size / 2);
+                lo_mask = nmask & lo_bitmask;
+                lo_mask |= lo_mask << (src.size / 2);
+                hi_mask = nmask & hi_bitmask;
+                hi_mask |= hi_mask >> (src.size / 2);
+            }
+            else
+            {
+                lo_mask = zip_lo(batch<T, A>(nmask), batch<T, A>(nmask)).to_native();
+                hi_mask = zip_hi(batch<T, A>(nmask), batch<T, A>(nmask)).to_native();
+            }
+            // }.
+            batch<T, A> src_lo = zip_lo(src.real(), src.imag());
+            batch<T, A> src_hi = zip_hi(src.real(), src.imag());
+            src_lo.store(reinterpret_cast<T*>(mem), lo_mask, mode);
+            src_hi.store(reinterpret_cast<T*>(mem) + src.size, hi_mask, mode);
+        }
+
         // transpose
         template <class A, class T>
         XSIMD_INLINE void transpose(batch<T, A>* matrix_begin, batch<T, A>* matrix_end, requires_arch<common>) noexcept

include/xsimd/types/xsimd_batch.hpp

@@ -1512,13 +1512,9 @@ namespace xsimd
 
     template <class T, class A>
     template <class Mode>
-    XSIMD_INLINE void batch<std::complex<T>, A>::store(value_type* mem, batch_bool<T, A> mask, Mode) const noexcept
+    XSIMD_INLINE void batch<std::complex<T>, A>::store(value_type* mem, batch_bool<T, A> mask, Mode mode) const noexcept
     {
-        alignas(A::alignment()) std::array<value_type, size> buffer;
-        store_aligned(buffer.data());
-        for (std::size_t i = 0; i < size; ++i)
-            if (mask.get(i))
-                mem[i] = buffer[i];
+        kernel::store_complex_masked<A>(mem, *this, mask, mode, A { });
     }
 
     template <class T, class A>