Eigen库对齐问题：declspec(align('16')) 的形参将不被对齐

一：错误提示：error C2719: '_Val': formal parameter with __declspec(align('16')) won't be aligned

英文提示:error C2719: 'p': formal parameter with __declspec(align('16')) won't be aligned

中文提示:error C2719: “p”: 具有 __declspec(align('16')) 的形参将不被对齐

导致整个现象的主要原因是使用了Eigen库，Eigen为了使用SSE加速，所以内存分配上使用了128位的指针。

更加准确的说法：

“First, "fixed-size" should be clear: an Eigen object
has fixed size if its number of rows and its number of columns are fixed at compile-time. So for example Matrix3f has fixed size, but MatrixXf doesn't (the opposite of fixed-size is dynamic-size).

The array of coefficients of a fixed-size Eigen object
is a plain "static array", it is not dynamically allocated. For example, the data behind a Matrix4f is just a "float array[16]".

Fixed-size objects are typically very small, which means that we want to handle them with zero runtime overhead -- both in terms of memory usage and of speed.

Now, vectorization (both SSE and AltiVec) works with 128-bit packets. Moreover, for performance reasons, these packets need to be have 128-bit alignment.

So it turns out that the only way that fixed-size Eigen objects
can be vectorized, is if their size is a multiple of 128 bits, or 16 bytes. Eigen will
then request 16-byte alignment for these objects, and henceforth rely on these objects being aligned so no runtime check for alignment is performed.”

二：解决方案：

分为四种情况：

• Cause 1: Structures having Eigen objects as members
• Cause 2: STL Containers
• Cause 3: Passing Eigen objects by value
• Cause 4: Compiler making a wrong assumption on stack alignment
  (for instance GCC on Windows)

每种情况可以对照官方的说法。可以参考如下链接

http://eigen.tuxfamily.org/dox/TopicUnalignedArrayAssert.html

在此不再重复表述。

三：我的解决方法：eigen 3.1.2有 bug,  重装Eigen；

关闭使用 预编译头；

关闭编译选项里面的16位对齐；

把Eigen::Vector4f    传值改为 传引用 Eigen::Vector4f &；

把结构体 传值 全部转换为 传引用。躲过对齐！！！

编译器选项：C/C++ ---代码生成---结构图成员对齐；

可恶：忘记了当初是怎么解决的！！！！！

四：对stl vector 进行修改：http://blog.csdn.net/pkueecser/article/details/8602656

产生问题的resize()方法，改为：

    void resize(size_type _Newsize, const _Ty & _Val)

    { // determine new length, padding with _Val elements as needed
        if (size() < _Newsize)

            _Insert_n(end(), _Newsize - size(), _Val);

        else if (_Newsize < size())

            erase(begin() + _Newsize, end());

    }

　　注意其中红色部分，将传值改为传参，避免在参数栈上创建被对齐的结构的对象。然后，在我们使用std::vector< Foo<5, float> >之前包含我们的foo_vector.hpp头文件，就可以正常使用了。  注意红色部分.....

是微软坑呢？还是Eigen坑？个人感觉微软更坑！！！

后记：

       添加 修改为 869行为引用之后，引发2562行的错误，修改方式为 把第二个参数0去掉。