From f2a94031c77de48559827492a8d1bd03c2ece1ed Mon Sep 17 00:00:00 2001
From: Dominique Belhachemi <db@domibel.de>
Date: Tue, 10 Nov 2026 21:25:41 +0000
Subject: [PATCH] Add LSX and LASX popcount kernels

LoongArch has a per-element population count.

vpcnt.{b,h,w,d} on LSX
xvpcnt.{b,h,w,d} on LASX
---
 include/xsimd/arch/xsimd_lasx.hpp | 14 ++++++++++++++
 include/xsimd/arch/xsimd_lsx.hpp  | 14 ++++++++++++++
 2 files changed, 28 insertions(+)

diff --git a/include/xsimd/arch/xsimd_lasx.hpp b/include/xsimd/arch/xsimd_lasx.hpp
index 0b10360fd..429e9c3fb 100644
--- a/include/xsimd/arch/xsimd_lasx.hpp
+++ b/include/xsimd/arch/xsimd_lasx.hpp
@@ -883,6 +883,20 @@ namespace xsimd
             return __lasx_xvfcmp_cun_d(self.data, self.data);
         }
 
+        // popcount
+        template <class A, class T, class = std::enable_if_t<std::is_integral_v<T>>>
+        XSIMD_INLINE batch<T, A> popcount(batch<T, A> const& self, requires_arch<lasx>) noexcept
+        {
+            if constexpr (sizeof(T) == 1)
+                return __lasx_xvpcnt_b(self.data);
+            else if constexpr (sizeof(T) == 2)
+                return __lasx_xvpcnt_h(self.data);
+            else if constexpr (sizeof(T) == 4)
+                return __lasx_xvpcnt_w(self.data);
+            else
+                return __lasx_xvpcnt_d(self.data);
+        }
+
         // select
         template <class A, class T>
         XSIMD_INLINE batch<T, A> select(batch_bool<T, A> const& cond, batch<T, A> const& true_br, batch<T, A> const& false_br, requires_arch<lasx>) noexcept
diff --git a/include/xsimd/arch/xsimd_lsx.hpp b/include/xsimd/arch/xsimd_lsx.hpp
index 7f1ca30fe..7ec0177a6 100644
--- a/include/xsimd/arch/xsimd_lsx.hpp
+++ b/include/xsimd/arch/xsimd_lsx.hpp
@@ -852,6 +852,20 @@ namespace xsimd
             return __lsx_vfcmp_cun_d(self.data, self.data);
         }
 
+        // popcount
+        template <class A, class T, class = std::enable_if_t<std::is_integral_v<T>>>
+        XSIMD_INLINE batch<T, A> popcount(batch<T, A> const& self, requires_arch<lsx>) noexcept
+        {
+            if constexpr (sizeof(T) == 1)
+                return __lsx_vpcnt_b(self.data);
+            else if constexpr (sizeof(T) == 2)
+                return __lsx_vpcnt_h(self.data);
+            else if constexpr (sizeof(T) == 4)
+                return __lsx_vpcnt_w(self.data);
+            else
+                return __lsx_vpcnt_d(self.data);
+        }
+
         // select
         template <class A, class T>
         XSIMD_INLINE batch<T, A> select(batch_bool<T, A> const& cond, batch<T, A> const& true_br, batch<T, A> const& false_br, requires_arch<lsx>) noexcept