Unable to access memory as SSE type on x86, but works fine on x64

Question

I have code written using MSVC SSE built-in functions.

__m128 zero = _mm_setzero_ps(); __m128 center = _mm_load_ps(&sphere.origin.x); __m128 boxmin = _mm_load_ps(&rhs.BottomLeftClosest.x); __m128 boxmax = _mm_load_ps(&rhs.TopRightFurthest.x); __m128 e = _mm_add_ps(_mm_max_ps(_mm_sub_ps(boxmin, center), zero), _mm_max_ps(_mm_sub_ps(center, boxmax), zero)); e = _mm_mul_ps(e, e); __declspec(align(16)) float arr[4]; _mm_store_ps(arr, e); float r = sphere.radius; return (arr[0] + arr[1] + arr[2] <= r * r); 

The type Math::Vector (which is the type sphere.origin , rhs.BottomLeftClosest and rhs.TopRightFurthest ), is actually an array of 3 floats. I have aligned them 16 bytes and this code works fine on x64. But on x86, I get an access violation by reading a null pointer. Any tips on where this comes from?

Answer 1

  __m128 center = _mm_load_ps(&sphere.origin.x); 

_mm_load_ps () requires the passed pointer to be aligned by 16 bytes. There is no evidence that you provided the correct alignment of the .origin.x sphere. You will need to use _mm_loadu_ps () if you cannot provide this guarantee.