在声明时初始化一个union数组

Question

我正在尝试在声明时初始化以下union数组：

typedef union { __m128d m;  float f[4]; } mat;
mat m[2] = { {{30467.14153,5910.1427,15846.23837,7271.22705},
{30467.14153,5910.1427,15846.23837,7271.22705}}};

但是我发现了以下错误：

matrix.c: In function ‘main’:
matrix.c:21: error: incompatible types in initialization
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[0]’)
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[0]’)
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[0]’)
matrix.c:21: error: incompatible types in initialization
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[1]’)
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[1]’)
matrix.c:21: warning: excess elements in union initializer
matrix.c:21: warning: (near initialization for ‘m[1]’)

Answer 1

引用此页面 ：

使用C89样式的初始值设定项，必须按声明的顺序初始化结构成员，并且只能初始化union的第一个成员。

所以，要么先将float数组放入，要么尽可能使用C99并写入：

mat m[2] = { { .f = { /* and so on */ } }, /* ... */ };

重要的是.f 。

Answer 2

您需要指明要初始化的联合字段。 尝试使用以下语法：

mat m[2] = {
    {.f = {30467.14153,5910.1427,15846.23837,7271.22705}},
    {.f = {30467.14153,5910.1427,15846.23837,7271.22705}}
};

这成功为我编译，没有任何警告。

Answer 3

尝试更改成员：

typedef union {
    float f[4];
    __m128d m;
} mat;
mat m[2] = { { {30467.14153,5910.1427,15846.23837,7271.22705},
               {30467.14153,5910.1427,15846.23837,7271.22705} } };

如果初始化union而没有成员规范，如.f = {...}，则初始化union的第一个成员。