Fixed kernel::store for booleans

xtensor-stack · Feb 24, 2025 · b3c882c · b3c882c
1 parent 339fade
commit b3c882c
Show file tree

Hide file tree

Showing 2 changed files with 30 additions and 11 deletions.
diff --git a/include/xsimd/arch/generic/xsimd_generic_memory.hpp b/include/xsimd/arch/generic/xsimd_generic_memory.hpp
@@ -514,7 +514,7 @@ namespace xsimd
         }
 
         // store
-        template <class T, class A>
+        template <class A, class T>
         XSIMD_INLINE void store(batch_bool<T, A> const& self, bool* mem, requires_arch<generic>) noexcept
         {
             using batch_type = batch<T, A>;

diff --git a/test/test_api.cpp b/test/test_api.cpp
@@ -12,6 +12,8 @@
 #include "xsimd/xsimd.hpp"
 #ifndef XSIMD_NO_SUPPORTED_ARCHITECTURE
 
+#include <functional>
+#include <numeric>
 #include <random>
 
 #include "test_utils.hpp"
@@ -20,6 +22,7 @@ template <class B>
 struct xsimd_api_test
 {
     using batch_type = B;
+    using batch_bool_type = typename B::batch_bool_type;
     using value_type = typename B::value_type;
     static constexpr size_t size = B::size;
     using array_type = std::array<value_type, size>;
@@ -81,17 +84,17 @@ struct xsimd_api_test
 
     void test_store()
     {
-        test_store_impl(i8_vec, "load int8_t");
-        test_store_impl(ui8_vec, "load uint8_t");
-        test_store_impl(i16_vec, "load int16_t");
-        test_store_impl(ui16_vec, "load uint16_t");
-        test_store_impl(i32_vec, "load int32_t");
-        test_store_impl(ui32_vec, "load uint32_t");
-        test_store_impl(i64_vec, "load int64_t");
-        test_store_impl(ui64_vec, "load uint64_t");
-        test_store_impl(f_vec, "load float");
+        test_store_impl(i8_vec, "store int8_t");
+        test_store_impl(ui8_vec, "store uint8_t");
+        test_store_impl(i16_vec, "store int16_t");
+        test_store_impl(ui16_vec, "store uint16_t");
+        test_store_impl(i32_vec, "store int32_t");
+        test_store_impl(ui32_vec, "store uint32_t");
+        test_store_impl(i64_vec, "store int64_t");
+        test_store_impl(ui64_vec, "store uint64_t");
+        test_store_impl(f_vec, "store float");
 #if XSIMD_WITH_NEON64 || !XSIMD_WITH_NEON
-        test_store_impl(d_vec, "load double");
+        test_store_impl(d_vec, "store double");
 #endif
     }
 
@@ -133,13 +136,29 @@ struct xsimd_api_test
         batch_type b = batch_type::load(v.data(), xsimd::aligned_mode());
         V res(size);
 
+        bool* b_data = new bool[size];
+
         xsimd::store_as(res.data(), b, xsimd::unaligned_mode());
         INFO(name, " unaligned");
         CHECK_VECTOR_EQ(res, v);
 
+        std::fill(b_data, b_data + size, false);
+        batch_bool_type bb = (b == b);
+        xsimd::store_as(b_data, bb, xsimd::unaligned_mode());
+        INFO(name, " batch_bool unaligned");
+        CHECK_UNARY(std::accumulate(b_data, b_data + size, true, std::logical_and<bool>()));
+
         xsimd::store_as(res.data(), b, xsimd::aligned_mode());
         INFO(name, " aligned");
         CHECK_VECTOR_EQ(res, v);
+
+        std::fill(b_data, b_data + size, false);
+        bb = (b == b);
+        xsimd::store_as(b_data, bb, xsimd::aligned_mode());
+        INFO(name, " batch_bool aligned");
+        CHECK_UNARY(std::accumulate(b_data, b_data + size, true, std::logical_and<bool>()));
+
+        delete[] b_data;
     }
 
     template <class T>