test.cpp

// One of many test.cpp I found. I think this one was for examining the generated assembly.
#include "xmmintrin.h"

#if 0
struct Vec3
{
	float x, y, z;
	__forceinline Vec3() {};
	__forceinline Vec3(float X, float Y, float Z) : x(X), y(Y), z(Z) {};
	__forceinline Vec3 operator*(const Vec3& t) const { return Vec3(x*t.x, y*t.y, z*t.z); };
	__forceinline Vec3 operator+(const Vec3& t) const { return Vec3(x+t.x, y+t.y, z+t.z); };
};

typedef const Vec3& Vec3Arg;
typedef Vec3 Vec3Res;
#endif

#if 1
struct Vec3
{
//	union
//	{
		__m128 vMM;
//		struct { float x, y, z; };
//	};

	__forceinline Vec3() {};
//	__forceinline Vec3(float X, float Y, float Z) : x(X), y(Y), z(Z) {};
	__forceinline Vec3(__m128 t) : vMM(t) {};

	__forceinline Vec3(Vec3& v) : vMM(v.vMM) {};
	__forceinline Vec3& operator=(Vec3& v) { vMM = v.vMM; return *this; };
	__forceinline Vec3 operator*(const Vec3& t) const { return Vec3(_mm_mul_ps(vMM, t.vMM)); };
	__forceinline Vec3 operator+(const Vec3& t) const { return Vec3(_mm_add_ps(vMM, t.vMM)); };
};

typedef const Vec3 Vec3Arg;
typedef const Vec3 Vec3Res;
#endif

#if 0
typedef __m128 Vec3;
typedef __m128 Vec3Arg;
typedef __m128 Vec3Res;
__forceinline Vec3Res operator+(Vec3Arg a, Vec3Arg b) { return _mm_add_ps(a,b); };
__forceinline Vec3Res operator*(Vec3Arg a, Vec3Arg b) { return _mm_mul_ps(a,b); };
#endif

__declspec(noinline) Vec3Res MulAdd(Vec3Arg a, Vec3Arg b, Vec3Arg c)
{
	return a * b + c;
}
// Dump of file test.obj

// File Type: COFF OBJECT
// 
// ?MulAdd@@YA?AUVec3@@ABU1@00@Z (struct Vec3 __cdecl MulAdd(struct Vec3 const &,struct Vec3 const &,struct Vec3 const &)):
//   00000000: 8B 44 24 08        mov         eax,dword ptr [esp+8]
//   00000004: 8B 4C 24 0C        mov         ecx,dword ptr [esp+0Ch]
//   00000008: F3 0F 10 00        movss       xmm0,dword ptr [eax]
//   0000000C: F3 0F 10 09        movss       xmm1,dword ptr [ecx]
//   00000010: F3 0F 10 51 04     movss       xmm2,dword ptr [ecx+4]
//   00000015: F3 0F 10 59 08     movss       xmm3,dword ptr [ecx+8]
//   0000001A: 8B 4C 24 10        mov         ecx,dword ptr [esp+10h]
//   0000001E: 0F 5A C0           cvtps2pd    xmm0,xmm0
//   00000021: 0F 5A C9           cvtps2pd    xmm1,xmm1
//   00000024: F2 0F 59 C1        mulsd       xmm0,xmm1
//   00000028: F3 0F 10 48 04     movss       xmm1,dword ptr [eax+4]
//   0000002D: 66 0F 5A C0        cvtpd2ps    xmm0,xmm0
//   00000031: 0F 5A D2           cvtps2pd    xmm2,xmm2
//   00000034: 0F 5A C9           cvtps2pd    xmm1,xmm1
//   00000037: F2 0F 59 CA        mulsd       xmm1,xmm2
//   0000003B: F3 0F 10 50 08     movss       xmm2,dword ptr [eax+8]
//   00000040: 8B 44 24 04        mov         eax,dword ptr [esp+4]
//   00000044: F3 0F 5A C0        cvtss2sd    xmm0,xmm0
//   00000048: 0F 5A DB           cvtps2pd    xmm3,xmm3
//   0000004B: 0F 5A D2           cvtps2pd    xmm2,xmm2
//   0000004E: F2 0F 59 D3        mulsd       xmm2,xmm3
//   00000052: F3 0F 10 19        movss       xmm3,dword ptr [ecx]
//   00000056: 0F 5A DB           cvtps2pd    xmm3,xmm3
//   00000059: F2 0F 58 D8        addsd       xmm3,xmm0
//   0000005D: 66 0F 5A C3        cvtpd2ps    xmm0,xmm3
//   00000061: F3 0F 11 00        movss       dword ptr [eax],xmm0
//   00000065: F3 0F 10 41 04     movss       xmm0,dword ptr [ecx+4]
//   0000006A: 0F 5A C0           cvtps2pd    xmm0,xmm0
//   0000006D: 66 0F 5A C9        cvtpd2ps    xmm1,xmm1
//   00000071: F3 0F 5A C9        cvtss2sd    xmm1,xmm1
//   00000075: F2 0F 58 C1        addsd       xmm0,xmm1
//   00000079: 66 0F 5A C0        cvtpd2ps    xmm0,xmm0
//   0000007D: F3 0F 11 40 04     movss       dword ptr [eax+4],xmm0
//   00000082: F3 0F 10 41 08     movss       xmm0,dword ptr [ecx+8]
//   00000087: 66 0F 5A D2        cvtpd2ps    xmm2,xmm2
//   0000008B: 0F 5A C0           cvtps2pd    xmm0,xmm0
//   0000008E: F3 0F 5A CA        cvtss2sd    xmm1,xmm2
//   00000092: F2 0F 58 C1        addsd       xmm0,xmm1
//   00000096: 66 0F 5A C0        cvtpd2ps    xmm0,xmm0
//   0000009A: F3 0F 11 40 08     movss       dword ptr [eax+8],xmm0
//   0000009F: C3                 ret

// File Type: COFF OBJECT
// 
// ?MulAdd@@YA?BUVec3@@ABU1@00@Z (struct Vec3 const __cdecl MulAdd(struct Vec3 const &,struct Vec3 const &,struct Vec3 const &)):
//   00000000: 55                 push        ebp
//   00000001: 8B EC              mov         ebp,esp
//   00000003: 83 E4 F0           and         esp,0FFFFFFF0h
//   00000006: 8B 4D 0C           mov         ecx,dword ptr [ebp+0Ch]
//   00000009: 0F 28 01           movaps      xmm0,xmmword ptr [ecx]
//   0000000C: 8B 55 10           mov         edx,dword ptr [ebp+10h]
//   0000000F: 0F 28 0A           movaps      xmm1,xmmword ptr [edx]
//   00000012: 8B 4D 14           mov         ecx,dword ptr [ebp+14h]
//   00000015: 8B 45 08           mov         eax,dword ptr [ebp+8]
//   00000018: 0F 59 C1           mulps       xmm0,xmm1
//   0000001B: 0F 28 09           movaps      xmm1,xmmword ptr [ecx]
//   0000001E: 0F 58 C1           addps       xmm0,xmm1
//   00000021: 0F 29 00           movaps      xmmword ptr [eax],xmm0
//   00000024: 8B E5              mov         esp,ebp
//   00000026: 5D                 pop         ebp
//   00000027: C3                 ret

// File Type: COFF OBJECT
// ?MulAdd@@YA?AT__m128@@T1@00@Z (union __m128 __cdecl MulAdd(union __m128,union __m128,union __m128)):
//   00000000: 0F 59 C1           mulps       xmm0,xmm1
//   00000003: 0F 58 C2           addps       xmm0,xmm2
//   00000006: C3                 ret