SSE_CrossProduct_Versions.cpp

	
	// SSE Cross Product Test Implementations Test - 
	temp_vec3<float> a(1.0f, 2.0f, 3.0f); temp_vec3<float> b(4.0f, 5.0f, 6.0f);

	for (std::size_t j = 0; j < 5; ++j)
	{
		std::cout << "\nITER = " << j << "\n";
		// Version A) \\

		auto start = std::chrono::system_clock::now();
		__m128 a_0 = _mm_shuffle_ps(a.sv, a.sv, _MM_SHUFFLE(3, 0, 2, 1)); // = a.y,a.z,a.x,0f 
		__m128 a_1 = _mm_shuffle_ps(a.sv, a.sv, _MM_SHUFFLE(3, 1, 0, 2)); // = a.z,a.x,a.y,0f
		__m128 b_0 = _mm_shuffle_ps(b.sv, b.sv, _MM_SHUFFLE(3, 1, 0, 2)); // = b.z,b.x,b.y,0f
		__m128 b_1 = _mm_shuffle_ps(b.sv, b.sv, _MM_SHUFFLE(3, 0, 2, 1)); // = b.y,b.z,b.x,0f
		// Actual CP Math -                                                          ( a_0      *     b_0  )      -    ( a_1        *     b_1 )
		__m128 res_A = _mm_sub_ps(_mm_mul_ps(a_0, b_0), _mm_mul_ps(a_1, b_1)); // (a.y,a.z,a.x,0f * b.z,b.x,b.y,0f) - (a.z,a.x,a.y,0f * b.y,b.z,b.x,0f)
		auto end = std::chrono::system_clock::now();
		//for (std::size_t i = 0; i < 3; ++i) std::cout << res_A.m128_f32[i] << " | ";
		std::cout << "\n ============== SSE Cross Product - Version A =============\n";
		std::cout << "Calc Time = " << std::chrono::duration_cast<std::chrono::nanoseconds>(end - start).count() << "ns\n";
		std::cout << "Calc Time = " << std::fixed << std::setprecision(8) << double((std::chrono::duration_cast<std::chrono::nanoseconds>(end - start)).count()) / double(1e+09) << "s\n";

		// Version B) \\

		start = std::chrono::system_clock::now();
		__m128 tmp0 = _mm_shuffle_ps(b.sv, b.sv, _MM_SHUFFLE(3, 0, 2, 1));
		// B --> tmp0 = (by,bz,bx,0)
		__m128 tmp1 = _mm_shuffle_ps(a.sv, a.sv, _MM_SHUFFLE(3, 0, 2, 1));
		// A -- > tmp1 = (ay,az,ax,0)
		tmp0 = _mm_mul_ps(tmp0, a.sv);
		// tmp0 =  (by * ax, bz * ay, bx * az, 0)
		tmp1 = _mm_mul_ps(tmp1, b.sv);
		// tmp1 =  (ay * bx, az * by, ax * bz, 0)
		__m128 tmp2 = _mm_sub_ps(tmp0, tmp1);
		// tmp2 =  ( ((by * ax) - (ay * bx)) , ((bz * ay) - (az * by)) , ((bx * az) - ax * bz), (0-0))
		__m128 res_B = _mm_shuffle_ps(tmp2, tmp2, _MM_SHUFFLE(3, 0, 2, 1)); // Have to shuffle to correct x,yz return order. 
		// res = ( ((bz * ay) - (az * by)), ((bx * az) - (ax * bz)), ((by * ax) - (ay * bx)),  (0-0))
		end = std::chrono::system_clock::now();
		// Debug __m128s. 
		//for (std::size_t j = 0; j < 3; ++j) std::cout << res_B.m128_f32[j] << " | ";	
		std::cout << "\n ============== SSE Cross Product - Version B =============\n";
		std::cout << "Calc Time = " << std::chrono::duration_cast<std::chrono::nanoseconds>(end - start).count() << "ns\n";
		std::cout << "Calc Time = " << std::fixed << std::setprecision(8) << double((std::chrono::duration_cast<std::chrono::nanoseconds>(end - start)).count()) / double(1e+09) << "s\n";
	}

// temp_vec3<T>
template <class T>
struct temp_vec3
{
	temp_vec3() = delete;
	temp_vec3(T xx, T yy, T zz) : x(xx), y(yy), z(zz) {}
	explicit temp_vec3(const T *a) { v = a };
	explicit temp_vec3(__m128 ssv) : sv(ssv) {}
	union
	{
		struct { T x, y, z; };
		T v[3];
		__m128 sv;
	};
};