SSE and Vector

[Tao]

hi.. Group,

I want to use __m128i with vector, but it seems does not work. The following
code generated a runtime error indicating that there is a memory corruption.
std::vector<__m128ilist;

__m128i x = _mm_set_epi32(0, 0, 0, 2);

list.push_back(x);

Can somebody tell me why this happen?

thanks.

[Tom Widmer [VC++ MVP]]

Tao wrote:

hi.. Group,

I want to use __m128i with vector, but it seems does not work. The following
code generated a runtime error indicating that there is a memory corruption.
std::vector<__m128ilist;

__m128i x = _mm_set_epi32(0, 0, 0, 2);

list.push_back(x);

Can somebody tell me why this happen?

__m128i requires 16-byte alignment, which vector doesn't guarantee to
provide (well, it should do, but __m128i has especially strict alignment
requirements - ones that aren't met by malloc). One way of working
around this would be to specify a special vector allocator which did
align the memory correctly. e.g. (VC2005 specific due to __alignof())

#include <vector>
#include <memory>
#include <malloc.h>
#include <emmintrin.h>

template <class T>
class aligned_allocator: public std::allocator<T>
{
public:
template <class U>
struct rebind
{
typedef aligned_allocator<Uother;
};

aligned_allocator(){}

template <class U>
aligned_allocator(aligned_allocator<Uconst& other){}

template <class U>
aligned_allocator& operator=(aligned_allocator<Uconst& other){
return *this;
}

T* allocate(std::size_t N, const void* hint)
{
return allocate(N);
}

T* allocate(size_type N)
{
return static_cast<T*>(_aligned_malloc(N * sizeof(T), __alignof(T)));
}

void deallocate(T* p, std::size_t N)
{
_aligned_free(p);
}
};

Then:
std::vector<__m128i, aligned_allocator<__m128i list;

Tom