Compiler Warning (Level 1) C4730

'main' : mixing _m64 and floating point expressions may result in incorrect code

A function uses __m64 and float/double types. Because the MMX and floating-point registers share the same physical register space (cannot be used simultaneously), using __m64 and float/double types in the same function can result in data corruption, possibly causing an exception.

To safely use __m64 types and floating-point types in the same function, each instruction that uses one of the types should be separated by the _m_empty() (for MMX) or _m_femms() (for 3DNow!) intrinsic.

The following sample generates C4730:

// C4730.cpp
// compile with: /W1
// processor: x86
#include "mmintrin.h"

void func(double)
{
}

int main(__m64 a, __m64 b)
{
   __m64 m;
   double f;
   f = 1.0;
   m = _m_paddb(a, b);
   // uncomment the next line to resolve C4730
   // _m_empty();
   func(f * 3.0);   // C4730
}