WIP

Author: serge-sans-paille

include/xsimd/arch/common/xsimd_common_math.hpp

@@ -2117,7 +2117,7 @@ namespace xsimd
             template <class Op, class A, class T>
             XSIMD_INLINE T reduce(Op, batch<T, A> const& self, std::integral_constant<unsigned, 1>) noexcept
             {
-              return ::xsimd::kernel::first(self, A{});
+                return ::xsimd::kernel::first(self, A {});
             }
 
             template <class Op, class A, class T, unsigned Lvl>

include/xsimd/arch/xsimd_rvv.hpp

@@ -1335,6 +1335,19 @@ namespace xsimd
             return result;
         }
 
+        // first
+        template <class A, class T, detail::rvv_enable_all_t<T> = 0>
+        XSIMD_INLINE T first(batch<T, A> const& arg, requires_arch<rvv>) noexcept
+        {
+            return detail::rvvmv_lane0(arg);
+        }
+
+        template <class A, class T, detail::rvv_enable_all_t<T> = 0>
+        XSIMD_INLINE std::complex<T> first(batch<std::complex<T>, A> const& arg, requires_arch<rvv>) noexcept
+        {
+            return std::complex<T> { detail::rvvmv_lane0(arg.real()), detail::rvvmv_lane0(arg.imag()) };
+        }
+
         // insert
         template <class A, class T, size_t I, detail::rvv_enable_all_t<T> = 0>
         XSIMD_INLINE batch<T, A> insert(batch<T, A> const& arg, T val, index<I>, requires_arch<rvv>) noexcept

include/xsimd/arch/xsimd_sse2.hpp

@@ -720,7 +720,6 @@ namespace xsimd
             }
         }
 
-
         // from_mask
         template <class A>
         XSIMD_INLINE batch_bool<float, A> from_mask(batch_bool<float, A> const&, uint64_t mask, requires_arch<sse2>) noexcept
@@ -1317,10 +1316,10 @@ namespace xsimd
             batch<T, A> step2 = _mm_shufflelo_epi16(acc1, mask2);
             batch<T, A> acc2 = max(acc1, step2);
             if (sizeof(T) == 2)
-                return first(acc2, A{});
+                return first(acc2, A {});
             batch<T, A> step3 = bitwise_cast<T>(bitwise_cast<uint16_t>(acc2) >> 8);
             batch<T, A> acc3 = max(acc2, step3);
-            return first(acc3, A{});
+            return first(acc3, A {});
         }
 
         // reduce_min
@@ -1339,10 +1338,10 @@ namespace xsimd
             batch<T, A> step2 = _mm_shufflelo_epi16(acc1, mask2);
             batch<T, A> acc2 = min(acc1, step2);
             if (sizeof(T) == 2)
-                return first(acc2, A{});
+                return first(acc2, A {});
             batch<T, A> step3 = bitwise_cast<T>(bitwise_cast<uint16_t>(acc2) >> 8);
             batch<T, A> acc3 = min(acc2, step3);
-            return first(acc3, A{});
+            return first(acc3, A {});
         }
 
         // rsqrt