Improved SIMD and swizzle operators interactions with GCC and Clang, Added raw SIMD API.

glm/CMakeLists.txt

@@ -19,6 +19,10 @@ file(GLOB_RECURSE GTX_SOURCE ./gtx/*.cpp)
 file(GLOB_RECURSE GTX_INLINE ./gtx/*.inl)
 file(GLOB_RECURSE GTX_HEADER ./gtx/*.hpp)
 
+file(GLOB_RECURSE SIMD_SOURCE ./simd/*.cpp)
+file(GLOB_RECURSE SIMD_INLINE ./simd/*.inl)
+file(GLOB_RECURSE SIMD_HEADER ./simd/*.h)
+
 source_group("Text Files" FILES ${ROOT_TEXT} ${ROOT_MD})
 source_group("Core Files" FILES ${CORE_SOURCE})
 source_group("Core Files" FILES ${CORE_INLINE})
@@ -29,6 +33,9 @@ source_group("GTC Files" FILES ${GTC_HEADER})
 source_group("GTX Files" FILES ${GTX_SOURCE})
 source_group("GTX Files" FILES ${GTX_INLINE})
 source_group("GTX Files" FILES ${GTX_HEADER})
+source_group("SIMD Files" FILES ${SIMD_SOURCE})
+source_group("SIMD Files" FILES ${SIMD_INLINE})
+source_group("SIMD Files" FILES ${SIMD_HEADER})
 
 include_directories(${CMAKE_CURRENT_SOURCE_DIR}/..)
 
@@ -37,7 +44,8 @@ if(GLM_TEST_ENABLE)
 		${ROOT_SOURCE}    ${ROOT_INLINE}    ${ROOT_HEADER}
 		${CORE_SOURCE}    ${CORE_INLINE}    ${CORE_HEADER}
 		${GTC_SOURCE}     ${GTC_INLINE}     ${GTC_HEADER}
-		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER})
+		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER}
+		${SIMD_SOURCE}    ${SIMD_INLINE}    ${SIMD_HEADER})
 endif(GLM_TEST_ENABLE)
 
 if(GLM_STATIC_LIBRARY_ENABLE)
@@ -45,7 +53,8 @@ if(GLM_STATIC_LIBRARY_ENABLE)
 		${ROOT_SOURCE}    ${ROOT_INLINE}    ${ROOT_HEADER}
 		${CORE_SOURCE}    ${CORE_INLINE}    ${CORE_HEADER}
 		${GTC_SOURCE}     ${GTC_INLINE}     ${GTC_HEADER}
-		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER})
+		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER}
+		${SIMD_SOURCE}    ${SIMD_INLINE}    ${SIMD_HEADER})
 endif(GLM_STATIC_LIBRARY_ENABLE)
 
 if(GLM_DYNAMIC_LIBRARY_ENABLE)
@@ -53,5 +62,6 @@ if(GLM_DYNAMIC_LIBRARY_ENABLE)
 		${ROOT_SOURCE}    ${ROOT_INLINE}    ${ROOT_HEADER}
 		${CORE_SOURCE}    ${CORE_INLINE}    ${CORE_HEADER}
 		${GTC_SOURCE}     ${GTC_INLINE}     ${GTC_HEADER}
-		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER})
+		${GTX_SOURCE}     ${GTX_INLINE}     ${GTX_HEADER}
+		${SIMD_SOURCE}    ${SIMD_INLINE}    ${SIMD_HEADER})
 endif(GLM_DYNAMIC_LIBRARY_ENABLE)

glm/detail/func_common.inl

@@ -734,6 +734,6 @@ namespace detail
 	}
 }//namespace glm
 
-#if GLM_ARCH != GLM_FORCE_PURE && GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#if GLM_ARCH != GLM_FORCE_PURE
 #	include "func_common_simd.inl"
 #endif

glm/detail/func_common_simd.inl

@@ -1,38 +1,6 @@
 namespace glm{
 namespace detail
 {
-#if(GLM_COMPILER & GLM_COMPILER_VC)
-#pragma warning(push)
-#pragma warning(disable : 4510 4512 4610)
-#endif
-
-	union ieee754_QNAN
-	{
-		const float f;
-		struct i
-		{
-			const unsigned int mantissa:23, exp:8, sign:1;
-		};
-
-		ieee754_QNAN() : f(0.0)/*, mantissa(0x7FFFFF), exp(0xFF), sign(0x0)*/ {}
-	};
-
-#if(GLM_COMPILER & GLM_COMPILER_VC)
-#pragma warning(pop)
-#endif
-
-	static const __m128 GLM_VAR_USED zero = _mm_setzero_ps();
-	static const __m128 GLM_VAR_USED one = _mm_set_ps1(1.0f);
-	static const __m128 GLM_VAR_USED minus_one = _mm_set_ps1(-1.0f);
-	static const __m128 GLM_VAR_USED two = _mm_set_ps1(2.0f);
-	static const __m128 GLM_VAR_USED three = _mm_set_ps1(3.0f);
-
-	static const ieee754_QNAN absMask;
-	static const __m128 GLM_VAR_USED abs4Mask = _mm_set_ps1(absMask.f);
-	static const __m128 GLM_VAR_USED _epi32_sign_mask = _mm_castsi128_ps(_mm_set1_epi32(static_cast<int>(0x80000000)));
-	static const __m128 GLM_VAR_USED _ps_2pow23 = _mm_set_ps1(8388608.0f);
-	static const __m128 GLM_VAR_USED _ps_1 = _mm_set_ps1(1.0f);
-
 /*
 	static const __m128 GLM_VAR_USED zero = _mm_setzero_ps();
 	static const __m128 GLM_VAR_USED one = _mm_set_ps1(1.0f);
@@ -137,151 +105,5 @@ namespace detail
 	static const __m128 GLM_VAR_USED _ps_log2_c0 = _mm_set_ps1(1.44269504088896340735992f);
 */
 
-	GLM_FUNC_QUALIFIER __m128 abs_ps(__m128 x)
-	{
-		return _mm_and_ps(abs4Mask, x);
-	}
-
-	//sign
-	GLM_FUNC_QUALIFIER __m128 sgn_ps(__m128 x)
-	{
-		__m128 Neg = _mm_set1_ps(-1.0f);
-		__m128 Pos = _mm_set1_ps(1.0f);
-
-		__m128 Cmp0 = _mm_cmplt_ps(x, zero);
-		__m128 Cmp1 = _mm_cmpgt_ps(x, zero);
-
-		__m128 And0 = _mm_and_ps(Cmp0, Neg);
-		__m128 And1 = _mm_and_ps(Cmp1, Pos);
-
-		return _mm_or_ps(And0, And1);
-	}
-
-	//round
-	GLM_FUNC_QUALIFIER __m128 rnd_ps(__m128 x)
-	{
-		__m128 and0 = _mm_and_ps(_epi32_sign_mask, x);
-		__m128 or0 = _mm_or_ps(and0, _ps_2pow23);
-		__m128 add0 = _mm_add_ps(x, or0);
-		__m128 sub0 = _mm_sub_ps(add0, or0);
-		return sub0;
-	}
-
-	//floor
-	GLM_FUNC_QUALIFIER __m128 flr_ps(__m128 x)
-	{
-		__m128 rnd0 = rnd_ps(x);
-		__m128 cmp0 = _mm_cmplt_ps(x, rnd0);
-		__m128 and0 = _mm_and_ps(cmp0, glm::detail::_ps_1);
-		__m128 sub0 = _mm_sub_ps(rnd0, and0);
-		return sub0;
-	}
-
-	//trunc
-	//GLM_FUNC_QUALIFIER __m128 _mm_trc_ps(__m128 v)
-	//{
-	//	return __m128();
-	//}
-
-	//roundEven
-	GLM_FUNC_QUALIFIER __m128 rde_ps(__m128 x)
-	{
-		__m128 and0 = _mm_and_ps(_epi32_sign_mask, x);
-		__m128 or0 = _mm_or_ps(and0, _ps_2pow23);
-		__m128 add0 = _mm_add_ps(x, or0);
-		__m128 sub0 = _mm_sub_ps(add0, or0);
-		return sub0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 ceil_ps(__m128 x)
-	{
-		__m128 rnd0 = rnd_ps(x);
-		__m128 cmp0 = _mm_cmpgt_ps(x, rnd0);
-		__m128 and0 = _mm_and_ps(cmp0, _ps_1);
-		__m128 add0 = _mm_add_ps(rnd0, and0);
-		return add0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 frc_ps(__m128 x)
-	{
-		__m128 flr0 = flr_ps(x);
-		__m128 sub0 = _mm_sub_ps(x, flr0);
-		return sub0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 mod_ps(__m128 x, __m128 y)
-	{
-		__m128 div0 = _mm_div_ps(x, y);
-		__m128 flr0 = flr_ps(div0);
-		__m128 mul0 = _mm_mul_ps(y, flr0);
-		__m128 sub0 = _mm_sub_ps(x, mul0);
-		return sub0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 clp_ps(__m128 v, __m128 minVal, __m128 maxVal)
-	{
-		__m128 min0 = _mm_min_ps(v, maxVal);
-		__m128 max0 = _mm_max_ps(min0, minVal);
-		return max0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 mix_ps(__m128 v1, __m128 v2, __m128 a)
-	{
-		__m128 sub0 = _mm_sub_ps(one, a);
-		__m128 mul0 = _mm_mul_ps(v1, sub0);
-		__m128 mul1 = _mm_mul_ps(v2, a);
-		__m128 add0 = _mm_add_ps(mul0, mul1);
-		return add0;
-	}
-
-	//step
-	GLM_FUNC_QUALIFIER __m128 stp_ps(__m128 edge, __m128 x)
-	{
-		__m128 cmp = _mm_cmple_ps(x, edge);
-		if(_mm_movemask_ps(cmp) == 0)
-			return one;
-		else
-			return zero;
-	}
-
-	// smoothstep
-	GLM_FUNC_QUALIFIER __m128 ssp_ps(__m128 edge0, __m128 edge1, __m128 x)
-	{
-		__m128 sub0 = _mm_sub_ps(x, edge0);
-		__m128 sub1 = _mm_sub_ps(edge1, edge0);
-		__m128 div0 = _mm_sub_ps(sub0, sub1);
-		__m128 clp0 = clp_ps(div0, zero, one);
-		__m128 mul0 = _mm_mul_ps(two, clp0);
-		__m128 sub2 = _mm_sub_ps(three, mul0);
-		__m128 mul1 = _mm_mul_ps(clp0, clp0);
-		__m128 mul2 = _mm_mul_ps(mul1, sub2);
-		return mul2;
-	}
-
-	/// \todo
-	//GLM_FUNC_QUALIFIER __m128 sse_nan_ps(__m128 x)
-	//{
-	//	__m128 empty;
-	//	return empty;
-	//}
-
-	/// \todo
-	//GLM_FUNC_QUALIFIER __m128 sse_inf_ps(__m128 x)
-	//{
-	//	__m128 empty;
-	//	return empty;
-	//}
-
-	// SSE scalar reciprocal sqrt using rsqrt op, plus one Newton-Rhaphson iteration
-	// By Elan Ruskin, http://assemblyrequired.crashworks.org/
-	GLM_FUNC_QUALIFIER __m128 sqrt_wip_ss(__m128 x)
-	{
-		__m128 const recip = _mm_rsqrt_ss(x);  // "estimate" opcode
-		__m128 const half = _mm_set_ps1(0.5f);
-		__m128 const halfrecip = _mm_mul_ss(half, recip);
-		__m128 const threeminus_xrr = _mm_sub_ss(three, _mm_mul_ss(x, _mm_mul_ss (recip, recip)));
-		return _mm_mul_ss(halfrecip, threeminus_xrr);
-	}
-
 }//namespace detail
 }//namespace glm

glm/detail/func_geometric.inl

@@ -171,6 +171,6 @@ namespace detail
 	}
 }//namespace glm
 
-#if GLM_ARCH != GLM_FORCE_PURE && GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#if GLM_ARCH != GLM_FORCE_PURE
 #	include "func_geometric_simd.inl"
 #endif

glm/detail/func_geometric_simd.inl

@@ -1,119 +1,22 @@
+#include "../simd/geometric.h"
+
 namespace glm{
 namespace detail
 {
-#if GLM_ARCH & GLM_ARCH_AVX
-	GLM_FUNC_QUALIFIER __m128 dot_ps(__m128 v1, __m128 v2)
-	{
-		return _mm_dp_ps(v1, v2, 0xff);
-	}
-#else
-	GLM_FUNC_QUALIFIER __m128 dot_ps(__m128 v1, __m128 v2)
-	{
-		__m128 mul0 = _mm_mul_ps(v1, v2);
-		__m128 swp0 = _mm_shuffle_ps(mul0, mul0, _MM_SHUFFLE(2, 3, 0, 1));
-		__m128 add0 = _mm_add_ps(mul0, swp0);
-		__m128 swp1 = _mm_shuffle_ps(add0, add0, _MM_SHUFFLE(0, 1, 2, 3));
-		__m128 add1 = _mm_add_ps(add0, swp1);
-		return add1;
-	}
-#endif
-
-	GLM_FUNC_QUALIFIER __m128 dot_ss(__m128 v1, __m128 v2)
-	{
-		__m128 mul0 = _mm_mul_ps(v1, v2);
-		__m128 mov0 = _mm_movehl_ps(mul0, mul0);
-		__m128 add0 = _mm_add_ps(mov0, mul0);
-		__m128 swp1 = _mm_shuffle_ps(add0, add0, 1);
-		__m128 add1 = _mm_add_ss(add0, swp1);
-		return add1;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 len_ps(__m128 x)
-	{
-		__m128 dot0 = dot_ps(x, x);
-		__m128 sqt0 = _mm_sqrt_ps(dot0);
-		return sqt0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 dst_ps(__m128 p0, __m128 p1)
-	{
-		__m128 sub0 = _mm_sub_ps(p0, p1);
-		__m128 len0 = len_ps(sub0);
-		return len0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 xpd_ps(__m128 v1, __m128 v2)
-	{
-		__m128 swp0 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 0, 2, 1));
-		__m128 swp1 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 1, 0, 2));
-		__m128 swp2 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 0, 2, 1));
-		__m128 swp3 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 1, 0, 2));
-		__m128 mul0 = _mm_mul_ps(swp0, swp3);
-		__m128 mul1 = _mm_mul_ps(swp1, swp2);
-		__m128 sub0 = _mm_sub_ps(mul0, mul1);
-		return sub0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 nrm_ps(__m128 v)
-	{
-		__m128 dot0 = dot_ps(v, v);
-		__m128 isr0 = _mm_rsqrt_ps(dot0);
-		__m128 mul0 = _mm_mul_ps(v, isr0);
-		return mul0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 ffd_ps(__m128 N, __m128 I, __m128 Nref)
-	{
-		__m128 dot0 = dot_ps(Nref, I);
-		__m128 sgn0 = sgn_ps(dot0);
-		__m128 mul0 = _mm_mul_ps(sgn0, glm::detail::minus_one);
-		__m128 mul1 = _mm_mul_ps(N, mul0);
-		return mul1;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 rfe_ps(__m128 I, __m128 N)
-	{
-		__m128 dot0 = dot_ps(N, I);
-		__m128 mul0 = _mm_mul_ps(N, dot0);
-		__m128 mul1 = _mm_mul_ps(mul0, glm::detail::two);
-		__m128 sub0 = _mm_sub_ps(I, mul1);
-		return sub0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 rfa_ps(__m128 I, __m128 N, __m128 eta)
-	{
-		__m128 dot0 = dot_ps(N, I);
-		__m128 mul0 = _mm_mul_ps(eta, eta);
-		__m128 mul1 = _mm_mul_ps(dot0, dot0);
-		__m128 sub0 = _mm_sub_ps(glm::detail::one, mul0);
-		__m128 sub1 = _mm_sub_ps(glm::detail::one, mul1);
-		__m128 mul2 = _mm_mul_ps(sub0, sub1);
-	
-		if(_mm_movemask_ps(_mm_cmplt_ss(mul2, glm::detail::zero)) == 0)
-			return glm::detail::zero;
-
-		__m128 sqt0 = _mm_sqrt_ps(mul2);
-		__m128 mul3 = _mm_mul_ps(eta, dot0);
-		__m128 add0 = _mm_add_ps(mul3, sqt0);
-		__m128 mul4 = _mm_mul_ps(add0, N);
-		__m128 mul5 = _mm_mul_ps(eta, I);
-		__m128 sub2 = _mm_sub_ps(mul5, mul4);
-
-		return sub2;
-	}
-
-	template <>
-	struct compute_dot<tvec4, float, simd>
-	{
-		GLM_FUNC_QUALIFIER static float call(tvec4<float, simd> const& x, tvec4<float, simd> const& y)
+#	if GLM_HAS_UNRESTRICTED_UNIONS
+		template <>
+		struct compute_dot<tvec4, float, simd>
 		{
-			__m128 const dot0 = dot_ss(x.data, y.data);
-
-			float Result = 0;
-			_mm_store_ss(&Result, dot0);
-			return Result;
-		}
-	};
+			GLM_FUNC_QUALIFIER static float call(tvec4<float, simd> const& x, tvec4<float, simd> const& y)
+			{
+				__m128 const dot0 = glm_dot_ss(x.data, y.data);
+
+				float Result = 0;
+				_mm_store_ss(&Result, dot0);
+				return Result;
+			}
+		};
+#	endif//GLM_HAS_UNRESTRICTED_UNIONS
 }//namespace detail
 }//namespace glm
 

glm/detail/func_integer.inl

@@ -359,7 +359,7 @@ namespace detail
 	}
 }//namespace glm
 
-#if GLM_ARCH != GLM_FORCE_PURE && GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#if GLM_ARCH != GLM_FORCE_PURE
 #	include "func_integer_simd.inl"
 #endif
 

glm/detail/func_matrix.inl

@@ -279,7 +279,7 @@ namespace detail
 	}
 }//namespace glm
 
-#if GLM_ARCH != GLM_FORCE_PURE && GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#if GLM_ARCH != GLM_FORCE_PURE
 #	include "func_matrix_simd.inl"
 #endif
 

glm/detail/func_matrix_simd.inl

@@ -3,1268 +3,19 @@
 
 #include "type_mat4x4.hpp"
 #include "func_geometric.hpp"
+#include "../simd/matrix.h"
 
-namespace glm{
-namespace detail
+namespace glm
 {
-	static const __m128 GLM_VAR_USED _m128_rad_ps = _mm_set_ps1(3.141592653589793238462643383279f / 180.f);
-	static const __m128 GLM_VAR_USED _m128_deg_ps = _mm_set_ps1(180.f / 3.141592653589793238462643383279f);
-
-	template <typename matType>
-	GLM_FUNC_QUALIFIER matType sse_comp_mul_ps
-	(
-		__m128 const in1[4],
-		__m128 const in2[4],
-		__m128 out[4]
-	)
-	{
-		out[0] = _mm_mul_ps(in1[0], in2[0]);
-		out[1] = _mm_mul_ps(in1[1], in2[1]);
-		out[2] = _mm_mul_ps(in1[2], in2[2]);
-		out[3] = _mm_mul_ps(in1[3], in2[3]);
-	}
-
-	GLM_FUNC_QUALIFIER void sse_add_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
-	{
-		{
-			out[0] = _mm_add_ps(in1[0], in2[0]);
-			out[1] = _mm_add_ps(in1[1], in2[1]);
-			out[2] = _mm_add_ps(in1[2], in2[2]);
-			out[3] = _mm_add_ps(in1[3], in2[3]);
-		}
-	}
-
-	GLM_FUNC_QUALIFIER void sse_sub_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
-	{
-		{
-			out[0] = _mm_sub_ps(in1[0], in2[0]);
-			out[1] = _mm_sub_ps(in1[1], in2[1]);
-			out[2] = _mm_sub_ps(in1[2], in2[2]);
-			out[3] = _mm_sub_ps(in1[3], in2[3]);
-		}
-	}
-
-	GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 const m[4], __m128 v)
-	{
-		__m128 v0 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 v1 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 v2 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 v3 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3));
-
-		__m128 m0 = _mm_mul_ps(m[0], v0);
-		__m128 m1 = _mm_mul_ps(m[1], v1);
-		__m128 m2 = _mm_mul_ps(m[2], v2);
-		__m128 m3 = _mm_mul_ps(m[3], v3);
-
-		__m128 a0 = _mm_add_ps(m0, m1);
-		__m128 a1 = _mm_add_ps(m2, m3);
-		__m128 a2 = _mm_add_ps(a0, a1);
-
-		return a2;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 v, __m128 const m[4])
-	{
-		__m128 i0 = m[0];
-		__m128 i1 = m[1];
-		__m128 i2 = m[2];
-		__m128 i3 = m[3];
-
-		__m128 m0 = _mm_mul_ps(v, i0);
-		__m128 m1 = _mm_mul_ps(v, i1);
-		__m128 m2 = _mm_mul_ps(v, i2);
-		__m128 m3 = _mm_mul_ps(v, i3);
-
-		__m128 u0 = _mm_unpacklo_ps(m0, m1);
-		__m128 u1 = _mm_unpackhi_ps(m0, m1);
-		__m128 a0 = _mm_add_ps(u0, u1);
-
-		__m128 u2 = _mm_unpacklo_ps(m2, m3);
-		__m128 u3 = _mm_unpackhi_ps(m2, m3);
-		__m128 a1 = _mm_add_ps(u2, u3);
-
-		__m128 f0 = _mm_movelh_ps(a0, a1);
-		__m128 f1 = _mm_movehl_ps(a1, a0);
-		__m128 f2 = _mm_add_ps(f0, f1);
-
-		return f2;
-	}
-
-	GLM_FUNC_QUALIFIER void sse_mul_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
-	{
-		{
-			__m128 e0 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 e1 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 e2 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 e3 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 m0 = _mm_mul_ps(in1[0], e0);
-			__m128 m1 = _mm_mul_ps(in1[1], e1);
-			__m128 m2 = _mm_mul_ps(in1[2], e2);
-			__m128 m3 = _mm_mul_ps(in1[3], e3);
-
-			__m128 a0 = _mm_add_ps(m0, m1);
-			__m128 a1 = _mm_add_ps(m2, m3);
-			__m128 a2 = _mm_add_ps(a0, a1);
-
-			out[0] = a2;
-		}
-
-		{
-			__m128 e0 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 e1 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 e2 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 e3 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 m0 = _mm_mul_ps(in1[0], e0);
-			__m128 m1 = _mm_mul_ps(in1[1], e1);
-			__m128 m2 = _mm_mul_ps(in1[2], e2);
-			__m128 m3 = _mm_mul_ps(in1[3], e3);
-
-			__m128 a0 = _mm_add_ps(m0, m1);
-			__m128 a1 = _mm_add_ps(m2, m3);
-			__m128 a2 = _mm_add_ps(a0, a1);
-
-			out[1] = a2;
-		}
-
-		{
-			__m128 e0 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 e1 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 e2 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 e3 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 m0 = _mm_mul_ps(in1[0], e0);
-			__m128 m1 = _mm_mul_ps(in1[1], e1);
-			__m128 m2 = _mm_mul_ps(in1[2], e2);
-			__m128 m3 = _mm_mul_ps(in1[3], e3);
-
-			__m128 a0 = _mm_add_ps(m0, m1);
-			__m128 a1 = _mm_add_ps(m2, m3);
-			__m128 a2 = _mm_add_ps(a0, a1);
-
-			out[2] = a2;
-		}
-
-		{
-			//(__m128&)_mm_shuffle_epi32(__m128i&)in2[0], _MM_SHUFFLE(3, 3, 3, 3))
-			__m128 e0 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 e1 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 e2 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 e3 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 m0 = _mm_mul_ps(in1[0], e0);
-			__m128 m1 = _mm_mul_ps(in1[1], e1);
-			__m128 m2 = _mm_mul_ps(in1[2], e2);
-			__m128 m3 = _mm_mul_ps(in1[3], e3);
-
-			__m128 a0 = _mm_add_ps(m0, m1);
-			__m128 a1 = _mm_add_ps(m2, m3);
-			__m128 a2 = _mm_add_ps(a0, a1);
-
-			out[3] = a2;
-		}
-	}
-
-	GLM_FUNC_QUALIFIER void sse_transpose_ps(__m128 const in[4], __m128 out[4])
-	{
-		__m128 tmp0 = _mm_shuffle_ps(in[0], in[1], 0x44);
-		__m128 tmp2 = _mm_shuffle_ps(in[0], in[1], 0xEE);
-		__m128 tmp1 = _mm_shuffle_ps(in[2], in[3], 0x44);
-		__m128 tmp3 = _mm_shuffle_ps(in[2], in[3], 0xEE);
-
-		out[0] = _mm_shuffle_ps(tmp0, tmp1, 0x88);
-		out[1] = _mm_shuffle_ps(tmp0, tmp1, 0xDD);
-		out[2] = _mm_shuffle_ps(tmp2, tmp3, 0x88);
-		out[3] = _mm_shuffle_ps(tmp2, tmp3, 0xDD);
-	}
-
-	GLM_FUNC_QUALIFIER __m128 sse_slow_det_ps(__m128 const in[4])
-	{
-		__m128 Fac0;
-		{
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
-			//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac0 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac1;
-		{
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
-			//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac1 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-
-		__m128 Fac2;
+#	if GLM_HAS_UNRESTRICTED_UNIONS
+		template <>
+		GLM_FUNC_QUALIFIER tmat4x4<float, simd> inverse(tmat4x4<float, simd> const& m)
 		{
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
-			//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac2 = _mm_sub_ps(Mul00, Mul01);
+			tmat4x4<float, simd> Result(uninitialize);
+			glm_inverse_ps(
+				*reinterpret_cast<__m128 const(*)[4]>(&m[0].data),
+				*reinterpret_cast<__m128(*)[4]>(&Result[0].data));
+			return Result;
 		}
-
-		__m128 Fac3;
-		{
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
-			//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac3 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac4;
-		{
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
-			//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac4 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac5;
-		{
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
-			//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac5 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
-		__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
-
-		// m[1][0]
-		// m[0][0]
-		// m[0][0]
-		// m[0][0]
-		__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][1]
-		// m[0][1]
-		// m[0][1]
-		// m[0][1]
-		__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][2]
-		// m[0][2]
-		// m[0][2]
-		// m[0][2]
-		__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][3]
-		// m[0][3]
-		// m[0][3]
-		// m[0][3]
-		__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
-		__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// col0
-		// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
-		// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
-		// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
-		// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
-		__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
-		__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
-		__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
-		__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
-		__m128 Add00 = _mm_add_ps(Sub00, Mul02);
-		__m128 Inv0 = _mm_mul_ps(SignB, Add00);
-
-		// col1
-		// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
-		// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
-		// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
-		// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
-		__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
-		__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
-		__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
-		__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
-		__m128 Add01 = _mm_add_ps(Sub01, Mul05);
-		__m128 Inv1 = _mm_mul_ps(SignA, Add01);
-
-		// col2
-		// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
-		// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
-		// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
-		// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
-		__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
-		__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
-		__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
-		__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
-		__m128 Add02 = _mm_add_ps(Sub02, Mul08);
-		__m128 Inv2 = _mm_mul_ps(SignB, Add02);
-
-		// col3
-		// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
-		// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
-		// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
-		// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
-		__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
-		__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
-		__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
-		__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
-		__m128 Add03 = _mm_add_ps(Sub03, Mul11);
-		__m128 Inv3 = _mm_mul_ps(SignA, Add03);
-
-		__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
-
-		//	valType Determinant = m[0][0] * Inverse[0][0]
-		//						+ m[0][1] * Inverse[1][0]
-		//						+ m[0][2] * Inverse[2][0]
-		//						+ m[0][3] * Inverse[3][0];
-		__m128 Det0 = dot_ps(in[0], Row2);
-		return Det0;
-	}
-
-	GLM_FUNC_QUALIFIER __m128 sse_detd_ps
-	(
-		__m128 const m[4]
-	)
-	{
-		// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(
-
-		//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-		//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-		//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-		//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-		//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-		//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-
-		// First 2 columns
- 		__m128 Swp2A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 1, 1, 2)));
- 		__m128 Swp3A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(3, 2, 3, 3)));
-		__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
-
-		// Second 2 columns
-		__m128 Swp2B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(3, 2, 3, 3)));
-		__m128 Swp3B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(0, 1, 1, 2)));
-		__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
-
-		// Columns subtraction
-		__m128 SubE = _mm_sub_ps(MulA, MulB);
-
-		// Last 2 rows
-		__m128 Swp2C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 0, 1, 2)));
-		__m128 Swp3C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(1, 2, 0, 0)));
-		__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
-		__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
-
-		//tvec4<T, P> DetCof(
-		//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
-		//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
-		//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
-		//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
-
-		__m128 SubFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubE), _MM_SHUFFLE(2, 1, 0, 0)));
-		__m128 SwpFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(0, 0, 0, 1)));
-		__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
-
-		__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
-		__m128 SubFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpB), _MM_SHUFFLE(3, 1, 1, 0)));//SubF[0], SubE[3], SubE[3], SubE[1];
-		__m128 SwpFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(1, 1, 2, 2)));
-		__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
-
-		__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
-
-		__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
-		__m128 SubFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpC), _MM_SHUFFLE(3, 3, 2, 0)));
-		__m128 SwpFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(2, 3, 3, 3)));
-		__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
-
-		__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
-		__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
-
-		//return m[0][0] * DetCof[0]
-		//	 + m[0][1] * DetCof[1]
-		//	 + m[0][2] * DetCof[2]
-		//	 + m[0][3] * DetCof[3];
-
-		return dot_ps(m[0], DetCof);
-	}
-
-	GLM_FUNC_QUALIFIER __m128 sse_det_ps
-	(
-		__m128 const m[4]
-	)
-	{
-		// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(add)
-
-		//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-		//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-		//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-		//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-		//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-		//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-
-		// First 2 columns
- 		__m128 Swp2A = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 1, 1, 2));
- 		__m128 Swp3A = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(3, 2, 3, 3));
-		__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
-
-		// Second 2 columns
-		__m128 Swp2B = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(3, 2, 3, 3));
-		__m128 Swp3B = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(0, 1, 1, 2));
-		__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
-
-		// Columns subtraction
-		__m128 SubE = _mm_sub_ps(MulA, MulB);
-
-		// Last 2 rows
-		__m128 Swp2C = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 0, 1, 2));
-		__m128 Swp3C = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(1, 2, 0, 0));
-		__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
-		__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
-
-		//tvec4<T, P> DetCof(
-		//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
-		//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
-		//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
-		//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
-
-		__m128 SubFacA = _mm_shuffle_ps(SubE, SubE, _MM_SHUFFLE(2, 1, 0, 0));
-		__m128 SwpFacA = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(0, 0, 0, 1));
-		__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
-
-		__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
-		__m128 SubFacB = _mm_shuffle_ps(SubTmpB, SubTmpB, _MM_SHUFFLE(3, 1, 1, 0));//SubF[0], SubE[3], SubE[3], SubE[1];
-		__m128 SwpFacB = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(1, 1, 2, 2));
-		__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
-
-		__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
-
-		__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
-		__m128 SubFacC = _mm_shuffle_ps(SubTmpC, SubTmpC, _MM_SHUFFLE(3, 3, 2, 0));
-		__m128 SwpFacC = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(2, 3, 3, 3));
-		__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
-
-		__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
-		__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
-
-		//return m[0][0] * DetCof[0]
-		//	 + m[0][1] * DetCof[1]
-		//	 + m[0][2] * DetCof[2]
-		//	 + m[0][3] * DetCof[3];
-
-		return dot_ps(m[0], DetCof);
-	}
-
-	GLM_FUNC_QUALIFIER void sse_inverse_ps(__m128 const in[4], __m128 out[4])
-	{
-		__m128 Fac0;
-		{
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
-			//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac0 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac1;
-		{
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
-			//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac1 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-
-		__m128 Fac2;
-		{
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
-			//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac2 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac3;
-		{
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
-			//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac3 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac4;
-		{
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
-			//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac4 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac5;
-		{
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
-			//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac5 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
-		__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
-
-		// m[1][0]
-		// m[0][0]
-		// m[0][0]
-		// m[0][0]
-		__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][1]
-		// m[0][1]
-		// m[0][1]
-		// m[0][1]
-		__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][2]
-		// m[0][2]
-		// m[0][2]
-		// m[0][2]
-		__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][3]
-		// m[0][3]
-		// m[0][3]
-		// m[0][3]
-		__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
-		__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// col0
-		// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
-		// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
-		// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
-		// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
-		__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
-		__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
-		__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
-		__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
-		__m128 Add00 = _mm_add_ps(Sub00, Mul02);
-		__m128 Inv0 = _mm_mul_ps(SignB, Add00);
-
-		// col1
-		// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
-		// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
-		// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
-		// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
-		__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
-		__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
-		__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
-		__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
-		__m128 Add01 = _mm_add_ps(Sub01, Mul05);
-		__m128 Inv1 = _mm_mul_ps(SignA, Add01);
-
-		// col2
-		// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
-		// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
-		// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
-		// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
-		__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
-		__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
-		__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
-		__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
-		__m128 Add02 = _mm_add_ps(Sub02, Mul08);
-		__m128 Inv2 = _mm_mul_ps(SignB, Add02);
-
-		// col3
-		// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
-		// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
-		// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
-		// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
-		__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
-		__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
-		__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
-		__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
-		__m128 Add03 = _mm_add_ps(Sub03, Mul11);
-		__m128 Inv3 = _mm_mul_ps(SignA, Add03);
-
-		__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
-
-		//	valType Determinant = m[0][0] * Inverse[0][0] 
-		//						+ m[0][1] * Inverse[1][0] 
-		//						+ m[0][2] * Inverse[2][0] 
-		//						+ m[0][3] * Inverse[3][0];
-		__m128 Det0 = dot_ps(in[0], Row2);
-		__m128 Rcp0 = _mm_div_ps(one, Det0);
-		//__m128 Rcp0 = _mm_rcp_ps(Det0);
-
-		//	Inverse /= Determinant;
-		out[0] = _mm_mul_ps(Inv0, Rcp0);
-		out[1] = _mm_mul_ps(Inv1, Rcp0);
-		out[2] = _mm_mul_ps(Inv2, Rcp0);
-		out[3] = _mm_mul_ps(Inv3, Rcp0);
-	}
-
-	GLM_FUNC_QUALIFIER void sse_inverse_fast_ps(__m128 const in[4], __m128 out[4])
-	{
-		__m128 Fac0;
-		{
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
-			//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac0 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac1;
-		{
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
-			//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac1 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-
-		__m128 Fac2;
-		{
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
-			//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac2 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac3;
-		{
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
-			//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac3 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac4;
-		{
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
-			//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac4 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac5;
-		{
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
-			//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
-
-			__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac5 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
-		__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
-
-		// m[1][0]
-		// m[0][0]
-		// m[0][0]
-		// m[0][0]
-		__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][1]
-		// m[0][1]
-		// m[0][1]
-		// m[0][1]
-		__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][2]
-		// m[0][2]
-		// m[0][2]
-		// m[0][2]
-		__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][3]
-		// m[0][3]
-		// m[0][3]
-		// m[0][3]
-		__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
-		__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// col0
-		// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
-		// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
-		// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
-		// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
-		__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
-		__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
-		__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
-		__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
-		__m128 Add00 = _mm_add_ps(Sub00, Mul02);
-		__m128 Inv0 = _mm_mul_ps(SignB, Add00);
-
-		// col1
-		// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
-		// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
-		// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
-		// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
-		__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
-		__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
-		__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
-		__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
-		__m128 Add01 = _mm_add_ps(Sub01, Mul05);
-		__m128 Inv1 = _mm_mul_ps(SignA, Add01);
-
-		// col2
-		// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
-		// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
-		// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
-		// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
-		__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
-		__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
-		__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
-		__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
-		__m128 Add02 = _mm_add_ps(Sub02, Mul08);
-		__m128 Inv2 = _mm_mul_ps(SignB, Add02);
-
-		// col3
-		// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
-		// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
-		// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
-		// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
-		__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
-		__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
-		__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
-		__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
-		__m128 Add03 = _mm_add_ps(Sub03, Mul11);
-		__m128 Inv3 = _mm_mul_ps(SignA, Add03);
-
-		__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
-
-		//	valType Determinant = m[0][0] * Inverse[0][0] 
-		//						+ m[0][1] * Inverse[1][0] 
-		//						+ m[0][2] * Inverse[2][0] 
-		//						+ m[0][3] * Inverse[3][0];
-		__m128 Det0 = dot_ps(in[0], Row2);
-		__m128 Rcp0 = _mm_rcp_ps(Det0);
-		//__m128 Rcp0 = _mm_div_ps(one, Det0);
-		//	Inverse /= Determinant;
-		out[0] = _mm_mul_ps(Inv0, Rcp0);
-		out[1] = _mm_mul_ps(Inv1, Rcp0);
-		out[2] = _mm_mul_ps(Inv2, Rcp0);
-		out[3] = _mm_mul_ps(Inv3, Rcp0);
-	}
-	/*
-	GLM_FUNC_QUALIFIER void sse_rotate_ps(__m128 const in[4], float Angle, float const v[3], __m128 out[4])
-	{
-		float a = glm::radians(Angle);
-		float c = cos(a);
-		float s = sin(a);
-
-		glm::vec4 AxisA(v[0], v[1], v[2], float(0));
-		__m128 AxisB = _mm_set_ps(AxisA.w, AxisA.z, AxisA.y, AxisA.x);
-		__m128 AxisC = detail::sse_nrm_ps(AxisB);
-
-		__m128 Cos0 = _mm_set_ss(c);
-		__m128 CosA = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 Sin0 = _mm_set_ss(s);
-		__m128 SinA = _mm_shuffle_ps(Sin0, Sin0, _MM_SHUFFLE(0, 0, 0, 0));
-
-		// tvec3<T, P> temp = (valType(1) - c) * axis;
-		__m128 Temp0 = _mm_sub_ps(one, CosA);
-		__m128 Temp1 = _mm_mul_ps(Temp0, AxisC);
-	
-		//Rotate[0][0] = c + temp[0] * axis[0];
-		//Rotate[0][1] = 0 + temp[0] * axis[1] + s * axis[2];
-		//Rotate[0][2] = 0 + temp[0] * axis[2] - s * axis[1];
-		__m128 Axis0 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 TmpA0 = _mm_mul_ps(Axis0, AxisC);
-		__m128 CosA0 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 1, 0));
-		__m128 TmpA1 = _mm_add_ps(CosA0, TmpA0);
-		__m128 SinA0 = SinA;//_mm_set_ps(0.0f, s, -s, 0.0f);
-		__m128 TmpA2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 1, 2, 3));
-		__m128 TmpA3 = _mm_mul_ps(SinA0, TmpA2);
-		__m128 TmpA4 = _mm_add_ps(TmpA1, TmpA3);
-
-		//Rotate[1][0] = 0 + temp[1] * axis[0] - s * axis[2];
-		//Rotate[1][1] = c + temp[1] * axis[1];
-		//Rotate[1][2] = 0 + temp[1] * axis[2] + s * axis[0];
-		__m128 Axis1 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 TmpB0 = _mm_mul_ps(Axis1, AxisC);
-		__m128 CosA1 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 0, 1));
-		__m128 TmpB1 = _mm_add_ps(CosA1, TmpB0);
-		__m128 SinB0 = SinA;//_mm_set_ps(-s, 0.0f, s, 0.0f);
-		__m128 TmpB2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 0, 3, 2));
-		__m128 TmpB3 = _mm_mul_ps(SinA0, TmpB2);
-		__m128 TmpB4 = _mm_add_ps(TmpB1, TmpB3);
-
-		//Rotate[2][0] = 0 + temp[2] * axis[0] + s * axis[1];
-		//Rotate[2][1] = 0 + temp[2] * axis[1] - s * axis[0];
-		//Rotate[2][2] = c + temp[2] * axis[2];
-		__m128 Axis2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 TmpC0 = _mm_mul_ps(Axis2, AxisC);
-		__m128 CosA2 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 0, 1, 1));
-		__m128 TmpC1 = _mm_add_ps(CosA2, TmpC0);
-		__m128 SinC0 = SinA;//_mm_set_ps(s, -s, 0.0f, 0.0f);
-		__m128 TmpC2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 3, 0, 1));
-		__m128 TmpC3 = _mm_mul_ps(SinA0, TmpC2);
-		__m128 TmpC4 = _mm_add_ps(TmpC1, TmpC3);
-
-		__m128 Result[4];
-		Result[0] = TmpA4;
-		Result[1] = TmpB4;
-		Result[2] = TmpC4;
-		Result[3] = _mm_set_ps(1, 0, 0, 0);
-
-		//tmat4x4<valType> Result(uninitialize);
-		//Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2];
-		//Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2];
-		//Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2];
-		//Result[3] = m[3];
-		//return Result;
-		sse_mul_ps(in, Result, out);
-	}
-	*/
-	GLM_FUNC_QUALIFIER void sse_outer_ps(__m128 const & c, __m128 const & r, __m128 out[4])
-	{
-		out[0] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(0, 0, 0, 0)));
-		out[1] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(1, 1, 1, 1)));
-		out[2] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(2, 2, 2, 2)));
-		out[3] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(3, 3, 3, 3)));
-	}
-}//namespace detail
-
-	template <>
-	GLM_FUNC_QUALIFIER tmat4x4<float, simd> inverse(tmat4x4<float, simd> const& m)
-	{
-		__m128 Fac0;
-		{
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
-			//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
-			//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac0 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac1;
-		{
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
-			//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
-			//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac1 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac2;
-		{
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
-			//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
-			//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac2 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac3;
-		{
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
-			//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
-			//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(3, 3, 3, 3));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(3, 3, 3, 3));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac3 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac4;
-		{
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
-			//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
-			//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(2, 2, 2, 2));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(2, 2, 2, 2));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac4 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 Fac5;
-		{
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
-			//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
-			//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
-
-			__m128 const Swp0a = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(1, 1, 1, 1));
-			__m128 const Swp0b = _mm_shuffle_ps(m[3].data, m[2].data, _MM_SHUFFLE(0, 0, 0, 0));
-
-			__m128 const Swp00 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(0, 0, 0, 0));
-			__m128 const Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
-			__m128 const Swp03 = _mm_shuffle_ps(m[2].data, m[1].data, _MM_SHUFFLE(1, 1, 1, 1));
-
-			__m128 const Mul00 = _mm_mul_ps(Swp00, Swp01);
-			__m128 const Mul01 = _mm_mul_ps(Swp02, Swp03);
-			Fac5 = _mm_sub_ps(Mul00, Mul01);
-		}
-
-		__m128 const SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
-		__m128 const SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
-
-		// m[1][0]
-		// m[0][0]
-		// m[0][0]
-		// m[0][0]
-		__m128 const Temp0 = _mm_shuffle_ps(m[1].data, m[0].data, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 const Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][1]
-		// m[0][1]
-		// m[0][1]
-		// m[0][1]
-		__m128 const Temp1 = _mm_shuffle_ps(m[1].data, m[0].data, _MM_SHUFFLE(1, 1, 1, 1));
-		__m128 const Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][2]
-		// m[0][2]
-		// m[0][2]
-		// m[0][2]
-		__m128 const Temp2 = _mm_shuffle_ps(m[1].data, m[0].data, _MM_SHUFFLE(2, 2, 2, 2));
-		__m128 const Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// m[1][3]
-		// m[0][3]
-		// m[0][3]
-		// m[0][3]
-		__m128 const Temp3 = _mm_shuffle_ps(m[1].data, m[0].data, _MM_SHUFFLE(3, 3, 3, 3));
-		__m128 const Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
-
-		// col0
-		// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
-		// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
-		// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
-		// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
-		__m128 const Mul00 = _mm_mul_ps(Vec1, Fac0);
-		__m128 const Mul01 = _mm_mul_ps(Vec2, Fac1);
-		__m128 const Mul02 = _mm_mul_ps(Vec3, Fac2);
-		__m128 const Sub00 = _mm_sub_ps(Mul00, Mul01);
-		__m128 const Add00 = _mm_add_ps(Sub00, Mul02);
-		__m128 const Inv0 = _mm_mul_ps(SignB, Add00);
-
-		// col1
-		// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
-		// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
-		// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
-		// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
-		__m128 const Mul03 = _mm_mul_ps(Vec0, Fac0);
-		__m128 const Mul04 = _mm_mul_ps(Vec2, Fac3);
-		__m128 const Mul05 = _mm_mul_ps(Vec3, Fac4);
-		__m128 const Sub01 = _mm_sub_ps(Mul03, Mul04);
-		__m128 const Add01 = _mm_add_ps(Sub01, Mul05);
-		__m128 const Inv1 = _mm_mul_ps(SignA, Add01);
-
-		// col2
-		// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
-		// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
-		// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
-		// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
-		__m128 const Mul06 = _mm_mul_ps(Vec0, Fac1);
-		__m128 const Mul07 = _mm_mul_ps(Vec1, Fac3);
-		__m128 const Mul08 = _mm_mul_ps(Vec3, Fac5);
-		__m128 const Sub02 = _mm_sub_ps(Mul06, Mul07);
-		__m128 const Add02 = _mm_add_ps(Sub02, Mul08);
-		__m128 const Inv2 = _mm_mul_ps(SignB, Add02);
-
-		// col3
-		// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
-		// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
-		// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
-		// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
-		__m128 const Mul09 = _mm_mul_ps(Vec0, Fac2);
-		__m128 const Mul10 = _mm_mul_ps(Vec1, Fac4);
-		__m128 const Mul11 = _mm_mul_ps(Vec2, Fac5);
-		__m128 const Sub03 = _mm_sub_ps(Mul09, Mul10);
-		__m128 const Add03 = _mm_add_ps(Sub03, Mul11);
-		__m128 const Inv3 = _mm_mul_ps(SignA, Add03);
-
-		__m128 const Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 const Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
-		__m128 const Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
-
-		//	valType Determinant = m[0][0] * Inverse[0][0] 
-		//						+ m[0][1] * Inverse[1][0] 
-		//						+ m[0][2] * Inverse[2][0] 
-		//						+ m[0][3] * Inverse[3][0];
-		__m128 const Det0 = detail::dot_ps(m[0].data, Row2);
-		__m128 const Rcp0 = _mm_rcp_ps(Det0);
-		//__m128 Rcp0 = _mm_div_ps(one, Det0);
-		//	Inverse /= Determinant;
-
-		tmat4x4<float, simd> Result(uninitialize);
-		Result[0].data = _mm_mul_ps(Inv0, Rcp0);
-		Result[1].data = _mm_mul_ps(Inv1, Rcp0);
-		Result[2].data = _mm_mul_ps(Inv2, Rcp0);
-		Result[3].data = _mm_mul_ps(Inv3, Rcp0);
-		return Result;
-	}
+#	endif// GLM_HAS_UNRESTRICTED_UNIONS
 }//namespace glm

glm/detail/setup.hpp

@@ -5,6 +5,7 @@
 
 #include <cassert>
 #include <cstddef>
+#include "../simd/platform.h"
 
 ///////////////////////////////////////////////////////////////////////////////////
 // Version
@@ -20,285 +21,6 @@
 #	pragma message ("GLM: version 0.9.8.0")
 #endif//GLM_MESSAGE
 
-///////////////////////////////////////////////////////////////////////////////////
-// Platform
-
-#define GLM_PLATFORM_UNKNOWN		0x00000000
-#define GLM_PLATFORM_WINDOWS		0x00010000
-#define GLM_PLATFORM_LINUX			0x00020000
-#define GLM_PLATFORM_APPLE			0x00040000
-//#define GLM_PLATFORM_IOS			0x00080000
-#define GLM_PLATFORM_ANDROID		0x00100000
-#define GLM_PLATFORM_CHROME_NACL	0x00200000
-#define GLM_PLATFORM_UNIX			0x00400000
-#define GLM_PLATFORM_QNXNTO			0x00800000
-#define GLM_PLATFORM_WINCE			0x01000000
-#define GLM_PLATFORM_CYGWIN			0x02000000
-
-#ifdef GLM_FORCE_PLATFORM_UNKNOWN
-#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
-#elif defined(__CYGWIN__)
-#	define GLM_PLATFORM GLM_PLATFORM_CYGWIN
-#elif defined(__QNXNTO__)
-#	define GLM_PLATFORM GLM_PLATFORM_QNXNTO
-#elif defined(__APPLE__)
-#	define GLM_PLATFORM GLM_PLATFORM_APPLE
-#elif defined(WINCE)
-#	define GLM_PLATFORM GLM_PLATFORM_WINCE
-#elif defined(_WIN32)
-#	define GLM_PLATFORM GLM_PLATFORM_WINDOWS
-#elif defined(__native_client__)
-#	define GLM_PLATFORM GLM_PLATFORM_CHROME_NACL
-#elif defined(__ANDROID__)
-#	define GLM_PLATFORM GLM_PLATFORM_ANDROID
-#elif defined(__linux)
-#	define GLM_PLATFORM GLM_PLATFORM_LINUX
-#elif defined(__unix)
-#	define GLM_PLATFORM GLM_PLATFORM_UNIX
-#else
-#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
-#endif//
-
-// Report platform detection
-#if(defined(GLM_MESSAGES) && !defined(GLM_MESSAGE_PLATFORM_DISPLAYED))
-#	define GLM_MESSAGE_PLATFORM_DISPLAYED
-#	if(GLM_PLATFORM & GLM_PLATFORM_QNXNTO)
-#		pragma message("GLM: QNX platform detected")
-//#	elif(GLM_PLATFORM & GLM_PLATFORM_IOS)
-//#		pragma message("GLM: iOS platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_APPLE)
-#		pragma message("GLM: Apple platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_WINCE)
-#		pragma message("GLM: WinCE platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_WINDOWS)
-#		pragma message("GLM: Windows platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_CHROME_NACL)
-#		pragma message("GLM: Native Client detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
-#		pragma message("GLM: Android platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_LINUX)
-#		pragma message("GLM: Linux platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_UNIX)
-#		pragma message("GLM: UNIX platform detected")
-#	elif(GLM_PLATFORM & GLM_PLATFORM_UNKNOWN)
-#		pragma message("GLM: platform unknown")
-#	else
-#		pragma message("GLM: platform not detected")
-#	endif
-#endif//GLM_MESSAGE
-
-///////////////////////////////////////////////////////////////////////////////////
-// Compiler
-
-// User defines: GLM_FORCE_COMPILER_UNKNOWN
-// TODO ? __llvm__
-
-#define GLM_COMPILER_UNKNOWN		0x00000000
-
-// Intel
-#define GLM_COMPILER_INTEL			0x00100000
-#define GLM_COMPILER_INTEL12		0x00100010
-#define GLM_COMPILER_INTEL12_1		0x00100020
-#define GLM_COMPILER_INTEL13		0x00100030
-#define GLM_COMPILER_INTEL14		0x00100040
-#define GLM_COMPILER_INTEL15		0x00100050
-#define GLM_COMPILER_INTEL16		0x00100060
-
-// Visual C++ defines
-#define GLM_COMPILER_VC				0x01000000
-#define GLM_COMPILER_VC2010			0x01000090
-#define GLM_COMPILER_VC2012			0x010000A0
-#define GLM_COMPILER_VC2013			0x010000B0
-#define GLM_COMPILER_VC2015			0x010000C0
-
-// GCC defines
-#define GLM_COMPILER_GCC			0x02000000
-#define GLM_COMPILER_GCC44			0x020000B0
-#define GLM_COMPILER_GCC45			0x020000C0
-#define GLM_COMPILER_GCC46			0x020000D0
-#define GLM_COMPILER_GCC47			0x020000E0
-#define GLM_COMPILER_GCC48			0x020000F0
-#define GLM_COMPILER_GCC49			0x02000100
-#define GLM_COMPILER_GCC50			0x02000200
-#define GLM_COMPILER_GCC51			0x02000300
-#define GLM_COMPILER_GCC52			0x02000400
-#define GLM_COMPILER_GCC53			0x02000500
-#define GLM_COMPILER_GCC60			0x02000600
-
-// CUDA
-#define GLM_COMPILER_CUDA			0x10000000
-#define GLM_COMPILER_CUDA40			0x10000040
-#define GLM_COMPILER_CUDA41			0x10000050
-#define GLM_COMPILER_CUDA42			0x10000060
-#define GLM_COMPILER_CUDA50			0x10000070
-#define GLM_COMPILER_CUDA60			0x10000080
-#define GLM_COMPILER_CUDA65			0x10000090
-#define GLM_COMPILER_CUDA70			0x100000A0
-#define GLM_COMPILER_CUDA75			0x100000B0
-
-// LLVM
-#define GLM_COMPILER_LLVM			0x20000000
-#define GLM_COMPILER_LLVM32			0x20000030
-#define GLM_COMPILER_LLVM33			0x20000040
-#define GLM_COMPILER_LLVM34			0x20000050
-#define GLM_COMPILER_LLVM35			0x20000060
-#define GLM_COMPILER_LLVM36			0x20000070
-#define GLM_COMPILER_LLVM37			0x20000080
-#define GLM_COMPILER_LLVM38			0x20000090
-#define GLM_COMPILER_LLVM39			0x200000A0
-
-// Apple Clang
-#define GLM_COMPILER_APPLE_CLANG	0x40000000
-#define GLM_COMPILER_APPLE_CLANG40	0x40000010
-#define GLM_COMPILER_APPLE_CLANG41	0x40000020
-#define GLM_COMPILER_APPLE_CLANG42	0x40000030
-#define GLM_COMPILER_APPLE_CLANG50	0x40000040
-#define GLM_COMPILER_APPLE_CLANG51	0x40000050
-#define GLM_COMPILER_APPLE_CLANG60	0x40000060
-#define GLM_COMPILER_APPLE_CLANG61	0x40000070
-
-// Build model
-#define GLM_MODEL_32				0x00000010
-#define GLM_MODEL_64				0x00000020
-
-// Force generic C++ compiler
-#ifdef GLM_FORCE_COMPILER_UNKNOWN
-#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
-
-#elif defined(__INTEL_COMPILER)
-#	if __INTEL_COMPILER == 1200
-#		define GLM_COMPILER GLM_COMPILER_INTEL12
-#	elif __INTEL_COMPILER == 1210
-#		define GLM_COMPILER GLM_COMPILER_INTEL12_1
-#	elif __INTEL_COMPILER == 1300
-#		define GLM_COMPILER GLM_COMPILER_INTEL13
-#	elif __INTEL_COMPILER == 1400
-#		define GLM_COMPILER GLM_COMPILER_INTEL14
-#	elif __INTEL_COMPILER == 1500
-#		define GLM_COMPILER GLM_COMPILER_INTEL15
-#	elif __INTEL_COMPILER >= 1600
-#		define GLM_COMPILER GLM_COMPILER_INTEL16
-#	else
-#		define GLM_COMPILER GLM_COMPILER_INTEL
-#	endif
-
-// CUDA
-#elif defined(__CUDACC__)
-#	if !defined(CUDA_VERSION) && !defined(GLM_FORCE_CUDA)
-#		include <cuda.h>  // make sure version is defined since nvcc does not define it itself!
-#	endif
-#	if CUDA_VERSION < 3000
-#		error "GLM requires CUDA 3.0 or higher"
-#	else
-#		define GLM_COMPILER GLM_COMPILER_CUDA
-#	endif
-
-// Visual C++
-#elif defined(_MSC_VER)
-#	if _MSC_VER < 1600
-#		error "GLM requires Visual C++ 2010 or higher"
-#	elif _MSC_VER == 1600
-#		define GLM_COMPILER GLM_COMPILER_VC2010
-#	elif _MSC_VER == 1700
-#		define GLM_COMPILER GLM_COMPILER_VC2012
-#	elif _MSC_VER == 1800
-#		define GLM_COMPILER GLM_COMPILER_VC2013
-#	elif _MSC_VER >= 1900
-#		define GLM_COMPILER GLM_COMPILER_VC2015
-#	else//_MSC_VER
-#		define GLM_COMPILER GLM_COMPILER_VC
-#	endif//_MSC_VER
-
-// Clang
-#elif defined(__clang__)
-#	if GLM_PLATFORM & GLM_PLATFORM_APPLE
-#		if __clang_major__ == 4 && __clang_minor__ == 0
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG40
-#		elif __clang_major__ == 4 && __clang_minor__ == 1
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG41
-#		elif __clang_major__ == 4 && __clang_minor__ == 2
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG42
-#		elif __clang_major__ == 5 && __clang_minor__ == 0
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG50
-#		elif __clang_major__ == 5 && __clang_minor__ == 1
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG51
-#		elif __clang_major__ == 6 && __clang_minor__ == 0
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG60
-#		elif __clang_major__ == 6 && __clang_minor__ >= 1
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG61
-#		elif __clang_major__ >= 7
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG61
-#		else
-#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG
-#		endif
-#	else
-#		if __clang_major__ == 3 && __clang_minor__ == 0
-#			define GLM_COMPILER GLM_COMPILER_LLVM30
-#		elif __clang_major__ == 3 && __clang_minor__ == 1
-#			define GLM_COMPILER GLM_COMPILER_LLVM31
-#		elif __clang_major__ == 3 && __clang_minor__ == 2
-#			define GLM_COMPILER GLM_COMPILER_LLVM32
-#		elif __clang_major__ == 3 && __clang_minor__ == 3
-#			define GLM_COMPILER GLM_COMPILER_LLVM33
-#		elif __clang_major__ == 3 && __clang_minor__ == 4
-#			define GLM_COMPILER GLM_COMPILER_LLVM34
-#		elif __clang_major__ == 3 && __clang_minor__ == 5
-#			define GLM_COMPILER GLM_COMPILER_LLVM35
-#		elif __clang_major__ == 3 && __clang_minor__ == 6
-#			define GLM_COMPILER GLM_COMPILER_LLVM36
-#		elif __clang_major__ == 3 && __clang_minor__ == 7
-#			define GLM_COMPILER GLM_COMPILER_LLVM37
-#		elif __clang_major__ == 3 && __clang_minor__ == 8
-#			define GLM_COMPILER GLM_COMPILER_LLVM38
-#		elif __clang_major__ == 3 && __clang_minor__ >= 9
-#			define GLM_COMPILER GLM_COMPILER_LLVM39
-#		elif __clang_major__ >= 4
-#			define GLM_COMPILER GLM_COMPILER_LLVM39
-#		else
-#			define GLM_COMPILER GLM_COMPILER_LLVM
-#		endif
-#	endif
-
-// G++
-#elif defined(__GNUC__) || defined(__MINGW32__)
-#	if (__GNUC__ == 4) && (__GNUC_MINOR__ == 2)
-#		define GLM_COMPILER (GLM_COMPILER_GCC42)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 3)
-#		define GLM_COMPILER (GLM_COMPILER_GCC43)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 4)
-#		define GLM_COMPILER (GLM_COMPILER_GCC44)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 5)
-#		define GLM_COMPILER (GLM_COMPILER_GCC45)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 6)
-#		define GLM_COMPILER (GLM_COMPILER_GCC46)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 7)
-#		define GLM_COMPILER (GLM_COMPILER_GCC47)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 8)
-#		define GLM_COMPILER (GLM_COMPILER_GCC48)
-#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ >= 9)
-#		define GLM_COMPILER (GLM_COMPILER_GCC49)
-#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 0)
-#		define GLM_COMPILER (GLM_COMPILER_GCC50)
-#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 1)
-#		define GLM_COMPILER (GLM_COMPILER_GCC51)
-#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 2)
-#		define GLM_COMPILER (GLM_COMPILER_GCC52)
-#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ >= 3)
-#		define GLM_COMPILER (GLM_COMPILER_GCC53)
-#	elif (__GNUC__ >= 6)
-#		define GLM_COMPILER (GLM_COMPILER_GCC60)
-#	else
-#		define GLM_COMPILER (GLM_COMPILER_GCC)
-#	endif
-
-#else
-#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
-#endif
-
-#ifndef GLM_COMPILER
-#error "GLM_COMPILER undefined, your compiler may not be supported by GLM. Add #define GLM_COMPILER 0 to ignore this message."
-#endif//GLM_COMPILER
-
 // Report compiler detection
 #if defined(GLM_MESSAGES) && !defined(GLM_MESSAGE_COMPILER_DISPLAYED)
 #	define GLM_MESSAGE_COMPILER_DISPLAYED
@@ -323,7 +45,7 @@
 // Build model
 
 #if defined(__arch64__) || defined(__LP64__) || defined(_M_X64) || defined(__ppc64__) || defined(__x86_64__)
-#		define GLM_MODEL	GLM_MODEL_64
+#	define GLM_MODEL	GLM_MODEL_64
 #elif defined(__i386__) || defined(__ppc__)
 #	define GLM_MODEL	GLM_MODEL_32
 #else
@@ -344,7 +66,7 @@
 #endif//GLM_MESSAGE
 
 ///////////////////////////////////////////////////////////////////////////////////
-// Platform
+// Instruction sets
 
 // User defines: GLM_FORCE_PURE GLM_FORCE_SSE2 GLM_FORCE_SSE3 GLM_FORCE_AVX GLM_FORCE_AVX2 GLM_FORCE_AVX2
 
@@ -689,16 +411,23 @@
 		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC2013))))
 #endif
 
-// N2544 Unrestricted unions
+// N2544 Unrestricted unions http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2008/n2544.pdf
+
+#define GLM_NOT_BUGGY_VC32BITS (!(GLM_MODEL == GLM_MODEL_32 && (GLM_COMPILER & GLM_COMPILER_VC) && GLM_COMPILER < GLM_COMPILER_VC2013))
+
 #if GLM_COMPILER & (GLM_COMPILER_LLVM | GLM_COMPILER_APPLE_CLANG)
 #	define GLM_HAS_UNRESTRICTED_UNIONS __has_feature(cxx_unrestricted_unions)
 #elif GLM_LANG & (GLM_LANG_CXX11_FLAG | GLM_LANG_CXXMS_FLAG)
 #	define GLM_HAS_UNRESTRICTED_UNIONS 1
 #else
 #	define GLM_HAS_UNRESTRICTED_UNIONS (GLM_LANG & GLM_LANG_CXX0X_FLAG) && (\
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_LANG & GLM_LANG_CXXMS_FLAG)) || \
+		((GLM_COMPILER & GLM_COMPILER_VC) && (GLM_COMPILER >= GLM_COMPILER_VC2015)) || \
 		((GLM_COMPILER & GLM_COMPILER_GCC) && (GLM_COMPILER >= GLM_COMPILER_GCC46)))
 #endif
 
+//#define GLM_HAS_ANONYMOUS_UNION (((GLM_LANG & GLM_LANG_CXXMS_FLAG) | (GLM_LANG & GLM_LANG_CXX11_FLAG)) && GLM_NOT_BUGGY_VC32BITS)
+
 // N2346
 #if GLM_COMPILER & (GLM_COMPILER_LLVM | GLM_COMPILER_APPLE_CLANG)
 #	define GLM_HAS_DEFAULTED_FUNCTIONS __has_feature(cxx_defaulted_functions)
@@ -803,9 +532,6 @@
 #	endif// GLM_COMPILER & GLM_COMPILER_VC
 #endif
 
-// Not standard
-#define GLM_HAS_ANONYMOUS_UNION (GLM_LANG & GLM_LANG_CXXMS_FLAG)
-
 ///////////////////////////////////////////////////////////////////////////////////
 // Static assert
 

glm/detail/type_mat4x4.inl

@@ -759,8 +759,6 @@ namespace detail
 	}
 }//namespace glm
 
-#if GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
-#if GLM_ARCH & GLM_ARCH_SSE2
-#	include "type_mat4x4_sse2.inl"
+#if GLM_ARCH != GLM_FORCE_PURE
+#	include "type_mat4x4_simd.inl"
 #endif
-#endif//

glm/detail/type_vec1.hpp

@@ -35,7 +35,7 @@
 #include "../fwd.hpp"
 #include "type_vec.hpp"
 #ifdef GLM_SWIZZLE
-#	if GLM_HAS_ANONYMOUS_UNION
+#	if GLM_HAS_UNRESTRICTED_UNIONS
 #		include "_swizzle.hpp"
 #	else
 #		include "_swizzle_func.hpp"
@@ -56,7 +56,7 @@ namespace glm
 
 		// -- Data --
 
-#		if GLM_HAS_ANONYMOUS_UNION
+#		if GLM_HAS_UNRESTRICTED_UNIONS
 			union
 			{
 				T x;
@@ -122,13 +122,13 @@ namespace glm
 
 		// -- Swizzle constructors --
 
-#		if(GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE))
+#		if(GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE))
 			template <int E0>
 			GLM_FUNC_DECL tvec1(detail::_swizzle<1, T, P, tvec1<T, P>, E0, -1,-2,-3> const & that)
 			{
 				*this = that();
 			}
-#		endif//(GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE))
+#		endif//(GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE))
 
 		// -- Unary arithmetic operators --
 

glm/detail/type_vec2.hpp

@@ -34,7 +34,7 @@
 
 #include "type_vec.hpp"
 #ifdef GLM_SWIZZLE
-#	if GLM_HAS_ANONYMOUS_UNION
+#	if GLM_HAS_UNRESTRICTED_UNIONS
 #		include "_swizzle.hpp"
 #	else
 #		include "_swizzle_func.hpp"
@@ -55,7 +55,7 @@ namespace glm
 
 		// -- Data --
 
-#		if GLM_HAS_ANONYMOUS_UNION
+#		if GLM_HAS_UNRESTRICTED_UNIONS
 			union
 			{
 				struct{ T x, y; };
@@ -128,13 +128,13 @@ namespace glm
 
 		// -- Swizzle constructors --
 
-#		if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 			template <int E0, int E1>
 			GLM_FUNC_DECL tvec2(detail::_swizzle<2, T, P, tvec2<T, P>, E0, E1,-1,-2> const & that)
 			{
 				*this = that();
 			}
-#		endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		endif// GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 		// -- Unary arithmetic operators --
 

glm/detail/type_vec3.hpp

@@ -34,7 +34,7 @@
 
 #include "type_vec.hpp"
 #ifdef GLM_SWIZZLE
-#	if GLM_HAS_ANONYMOUS_UNION
+#	if GLM_HAS_UNRESTRICTED_UNIONS
 #		include "_swizzle.hpp"
 #	else
 #		include "_swizzle_func.hpp"
@@ -55,7 +55,7 @@ namespace glm
 
 		// -- Data --
 
-#		if GLM_HAS_ANONYMOUS_UNION
+#		if GLM_HAS_UNRESTRICTED_UNIONS
 			union
 			{
 				struct{ T x, y, z; };
@@ -138,7 +138,7 @@ namespace glm
 
 		// -- Swizzle constructors --
 
-#		if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 			template <int E0, int E1, int E2>
 			GLM_FUNC_DECL tvec3(detail::_swizzle<3, T, P, tvec3<T, P>, E0, E1, E2, -1> const & that)
 			{
@@ -156,7 +156,7 @@ namespace glm
 			{
 				*this = tvec3<T, P>(scalar, v());
 			}
-#		endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		endif// GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 		// -- Unary arithmetic operators --
 

glm/detail/type_vec4.hpp

@@ -52,9 +52,7 @@ namespace detail
 		typedef T type[4];
 	};
 
-#	define GLM_NOT_BUGGY_VC32BITS !(GLM_MODEL == GLM_MODEL_32 && (GLM_COMPILER & GLM_COMPILER_VC) && GLM_COMPILER < GLM_COMPILER_VC2013)
-
-#	if (GLM_ARCH & GLM_ARCH_SSE2) && GLM_NOT_BUGGY_VC32BITS
+#	if (GLM_ARCH & GLM_ARCH_SSE2)
 		template <>
 		struct simd_data<float, simd>
 		{
@@ -74,7 +72,7 @@ namespace detail
 		};
 #	endif
 
-#	if (GLM_ARCH & GLM_ARCH_AVX) && GLM_NOT_BUGGY_VC32BITS
+#	if (GLM_ARCH & GLM_ARCH_AVX)
 		template <>
 		struct simd_data<double, simd>
 		{
@@ -82,7 +80,7 @@ namespace detail
 		};
 #	endif
 
-#	if (GLM_ARCH & GLM_ARCH_AVX2) && GLM_NOT_BUGGY_VC32BITS
+#	if (GLM_ARCH & GLM_ARCH_AVX2)
 		template <>
 		struct simd_data<int64, simd>
 		{
@@ -109,7 +107,7 @@ namespace detail
 
 		// -- Data --
 
-#		if GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#		if GLM_HAS_UNRESTRICTED_UNIONS
 			union
 			{
 				struct { T x, y, z, w;};
@@ -213,7 +211,7 @@ namespace detail
 
 		// -- Swizzle constructors --
 
-#		if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 			template <int E0, int E1, int E2, int E3>
 			GLM_FUNC_DECL tvec4(detail::_swizzle<4, T, P, tvec4<T, P>, E0, E1, E2, E3> const & that)
 			{
@@ -255,7 +253,7 @@ namespace detail
 			{
 				*this = tvec4<T, P>(x, v());
 			}
-#		endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#		endif// GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 		// -- Unary arithmetic operators --
 

glm/detail/type_vec4.inl

@@ -1180,6 +1180,6 @@ namespace glm
 	}
 }//namespace glm
 
-#if GLM_ARCH != GLM_FORCE_PURE && GLM_HAS_ANONYMOUS_UNION && GLM_NOT_BUGGY_VC32BITS
+#if GLM_ARCH != GLM_FORCE_PURE
 #	include "type_vec4_simd.inl"
 #endif

glm/detail/type_vec4_simd.inl

@@ -32,6 +32,8 @@
 
 namespace glm
 {
+#	if GLM_HAS_UNRESTRICTED_UNIONS
+
 #	if !GLM_HAS_DEFAULTED_FUNCTIONS
 		template <>
 		GLM_FUNC_QUALIFIER tvec4<float, simd>::tvec4()
@@ -90,4 +92,5 @@ namespace glm
 		Result.data = _mm_mul_ps(v1.data, v2.data);
 		return Result;
 	}
+#endif//GLM_HAS_UNRESTRICTED_UNIONS
 }//namespace glm

glm/simd/common.h

@@ -0,0 +1,169 @@
+#if(GLM_COMPILER & GLM_COMPILER_VC)
+#pragma warning(push)
+#pragma warning(disable : 4510 4512 4610)
+#endif
+
+	union ieee754_QNAN
+	{
+		const float f;
+		struct i
+		{
+			const unsigned int mantissa:23, exp:8, sign:1;
+		};
+
+		ieee754_QNAN() : f(0.0)/*, mantissa(0x7FFFFF), exp(0xFF), sign(0x0)*/ {}
+	};
+
+#if(GLM_COMPILER & GLM_COMPILER_VC)
+#pragma warning(pop)
+#endif
+
+static const __m128 GLM_VAR_USED glm_zero = _mm_setzero_ps();
+static const __m128 GLM_VAR_USED glm_one = _mm_set_ps1(1.0f);
+static const __m128 GLM_VAR_USED glm_half = _mm_set_ps1(0.5f);
+static const __m128 GLM_VAR_USED glm_minus_one = _mm_set_ps1(-1.0f);
+static const __m128 GLM_VAR_USED glm_two = _mm_set_ps1(2.0f);
+static const __m128 GLM_VAR_USED glm_three = _mm_set_ps1(3.0f);
+
+static const ieee754_QNAN glm_abs_mask;
+static const __m128 GLM_VAR_USED glm_abs4_mask = _mm_set_ps1(glm_abs_mask.f);
+static const __m128 GLM_VAR_USED glm_epi32_sign_mask = _mm_castsi128_ps(_mm_set1_epi32(static_cast<int>(0x80000000)));
+static const __m128 GLM_VAR_USED glm_ps_2pow23 = _mm_set_ps1(8388608.0f);
+static const __m128 GLM_VAR_USED glm_ps_1 = _mm_set_ps1(1.0f);
+
+GLM_FUNC_QUALIFIER __m128 glm_abs_ps(__m128 x)
+{
+	return _mm_and_ps(glm_abs4_mask, x);
+}
+
+//sign
+GLM_FUNC_QUALIFIER __m128 glm_sgn_ps(__m128 x)
+{
+	__m128 const Cmp0 = _mm_cmplt_ps(x, glm_zero);
+	__m128 const Cmp1 = _mm_cmpgt_ps(x, glm_zero);
+	__m128 const And0 = _mm_and_ps(Cmp0, glm_minus_one);
+	__m128 const And1 = _mm_and_ps(Cmp1, glm_one);
+	return _mm_or_ps(And0, And1);
+}
+
+//round
+GLM_FUNC_QUALIFIER __m128 glm_rnd_ps(__m128 x)
+{
+	__m128 const and0 = _mm_and_ps(glm_epi32_sign_mask, x);
+	__m128 const or0 = _mm_or_ps(and0, glm_ps_2pow23);
+	__m128 const add0 = _mm_add_ps(x, or0);
+	__m128 const sub0 = _mm_sub_ps(add0, or0);
+	return sub0;
+}
+
+//floor
+GLM_FUNC_QUALIFIER __m128 glm_flr_ps(__m128 x)
+{
+	__m128 const rnd0 = glm_rnd_ps(x);
+	__m128 const cmp0 = _mm_cmplt_ps(x, rnd0);
+	__m128 const and0 = _mm_and_ps(cmp0, glm_ps_1);
+	__m128 const sub0 = _mm_sub_ps(rnd0, and0);
+	return sub0;
+}
+
+//trunc
+//GLM_FUNC_QUALIFIER __m128 _mm_trc_ps(__m128 v)
+//{
+//	return __m128();
+//}
+
+//roundEven
+GLM_FUNC_QUALIFIER __m128 glm_rde_ps(__m128 x)
+{
+	__m128 const and0 = _mm_and_ps(glm_epi32_sign_mask, x);
+	__m128 const or0 = _mm_or_ps(and0, glm_ps_2pow23);
+	__m128 const add0 = _mm_add_ps(x, or0);
+	__m128 const sub0 = _mm_sub_ps(add0, or0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_ceil_ps(__m128 x)
+{
+	__m128 const rnd0 = glm_rnd_ps(x);
+	__m128 const cmp0 = _mm_cmpgt_ps(x, rnd0);
+	__m128 const and0 = _mm_and_ps(cmp0, glm_ps_1);
+	__m128 const add0 = _mm_add_ps(rnd0, and0);
+	return add0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_frc_ps(__m128 x)
+{
+	__m128 const flr0 = glm_flr_ps(x);
+	__m128 const sub0 = _mm_sub_ps(x, flr0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_mod_ps(__m128 x, __m128 y)
+{
+	__m128 const div0 = _mm_div_ps(x, y);
+	__m128 const flr0 = glm_flr_ps(div0);
+	__m128 const mul0 = _mm_mul_ps(y, flr0);
+	__m128 const sub0 = _mm_sub_ps(x, mul0);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_clp_ps(__m128 v, __m128 minVal, __m128 maxVal)
+{
+	__m128 const min0 = _mm_min_ps(v, maxVal);
+	__m128 const max0 = _mm_max_ps(min0, minVal);
+	return max0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_mix_ps(__m128 v1, __m128 v2, __m128 a)
+{
+	__m128 const sub0 = _mm_sub_ps(glm_one, a);
+	__m128 const mul0 = _mm_mul_ps(v1, sub0);
+	__m128 const mul1 = _mm_mul_ps(v2, a);
+	__m128 const add0 = _mm_add_ps(mul0, mul1);
+	return add0;
+}
+
+//step
+GLM_FUNC_QUALIFIER __m128 glm_stp_ps(__m128 edge, __m128 x)
+{
+	__m128 const cmp = _mm_cmple_ps(x, edge);
+	return _mm_movemask_ps(cmp) == 0 ? glm_one : glm_zero;
+}
+
+// smoothstep
+GLM_FUNC_QUALIFIER __m128 glm_ssp_ps(__m128 edge0, __m128 edge1, __m128 x)
+{
+	__m128 const sub0 = _mm_sub_ps(x, edge0);
+	__m128 const sub1 = _mm_sub_ps(edge1, edge0);
+	__m128 const div0 = _mm_sub_ps(sub0, sub1);
+	__m128 const clp0 = glm_clp_ps(div0, glm_zero, glm_one);
+	__m128 const mul0 = _mm_mul_ps(glm_two, clp0);
+	__m128 const sub2 = _mm_sub_ps(glm_three, mul0);
+	__m128 const mul1 = _mm_mul_ps(clp0, clp0);
+	__m128 const mul2 = _mm_mul_ps(mul1, sub2);
+	return mul2;
+}
+
+/// \todo
+//GLM_FUNC_QUALIFIER __m128 glm_nan_ps(__m128 x)
+//{
+//	__m128 empty;
+//	return empty;
+//}
+
+/// \todo
+//GLM_FUNC_QUALIFIER __m128 glm_inf_ps(__m128 x)
+//{
+//	__m128 empty;
+//	return empty;
+//}
+
+// SSE scalar reciprocal sqrt using rsqrt op, plus one Newton-Rhaphson iteration
+// By Elan Ruskin, http://assemblyrequired.crashworks.org/
+GLM_FUNC_QUALIFIER __m128 glm_sqrt_wip_ss(__m128 x)
+{
+	__m128 const recip = _mm_rsqrt_ss(x);  // "estimate" opcode
+	__m128 const halfrecip = _mm_mul_ss(glm_half, recip);
+	__m128 const threeminus_xrr = _mm_sub_ss(glm_three, _mm_mul_ss(x, _mm_mul_ss(recip, recip)));
+	return _mm_mul_ss(halfrecip, threeminus_xrr);
+}

glm/simd/geometric.h

@@ -0,0 +1,101 @@
+#pragma once
+
+#include "common.h"
+
+GLM_FUNC_QUALIFIER __m128 glm_dot_ps(__m128 v1, __m128 v2)
+{
+#	if GLM_ARCH & GLM_ARCH_AVX
+		return _mm_dp_ps(v1, v2, 0xff);
+#	else
+		__m128 const mul0 = _mm_mul_ps(v1, v2);
+		__m128 const swp0 = _mm_shuffle_ps(mul0, mul0, _MM_SHUFFLE(2, 3, 0, 1));
+		__m128 const add0 = _mm_add_ps(mul0, swp0);
+		__m128 const swp1 = _mm_shuffle_ps(add0, add0, _MM_SHUFFLE(0, 1, 2, 3));
+		__m128 const add1 = _mm_add_ps(add0, swp1);
+		return add1;
+#	endif
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_dot_ss(__m128 v1, __m128 v2)
+{
+	__m128 const mul0 = _mm_mul_ps(v1, v2);
+	__m128 const mov0 = _mm_movehl_ps(mul0, mul0);
+	__m128 const add0 = _mm_add_ps(mov0, mul0);
+	__m128 const swp1 = _mm_shuffle_ps(add0, add0, 1);
+	__m128 const add1 = _mm_add_ss(add0, swp1);
+	return add1;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_len_ps(__m128 x)
+{
+	__m128 const dot0 = glm_dot_ps(x, x);
+	__m128 const sqt0 = _mm_sqrt_ps(dot0);
+	return sqt0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_dst_ps(__m128 p0, __m128 p1)
+{
+	__m128 sub0 = _mm_sub_ps(p0, p1);
+	__m128 len0 = glm_len_ps(sub0);
+	return len0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_xpd_ps(__m128 v1, __m128 v2)
+{
+	__m128 swp0 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 0, 2, 1));
+	__m128 swp1 = _mm_shuffle_ps(v1, v1, _MM_SHUFFLE(3, 1, 0, 2));
+	__m128 swp2 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 0, 2, 1));
+	__m128 swp3 = _mm_shuffle_ps(v2, v2, _MM_SHUFFLE(3, 1, 0, 2));
+	__m128 mul0 = _mm_mul_ps(swp0, swp3);
+	__m128 mul1 = _mm_mul_ps(swp1, swp2);
+	__m128 sub0 = _mm_sub_ps(mul0, mul1);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_nrm_ps(__m128 v)
+{
+	__m128 dot0 = glm_dot_ps(v, v);
+	__m128 isr0 = _mm_rsqrt_ps(dot0);
+	__m128 mul0 = _mm_mul_ps(v, isr0);
+	return mul0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_ffd_ps(__m128 N, __m128 I, __m128 Nref)
+{
+	__m128 dot0 = glm_dot_ps(Nref, I);
+	__m128 sgn0 = glm_sgn_ps(dot0);
+	__m128 mul0 = _mm_mul_ps(sgn0, glm_minus_one);
+	__m128 mul1 = _mm_mul_ps(N, mul0);
+	return mul1;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_rfe_ps(__m128 I, __m128 N)
+{
+	__m128 dot0 = glm_dot_ps(N, I);
+	__m128 mul0 = _mm_mul_ps(N, dot0);
+	__m128 mul1 = _mm_mul_ps(mul0, glm_two);
+	__m128 sub0 = _mm_sub_ps(I, mul1);
+	return sub0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_rfa_ps(__m128 I, __m128 N, __m128 eta)
+{
+	__m128 dot0 = glm_dot_ps(N, I);
+	__m128 mul0 = _mm_mul_ps(eta, eta);
+	__m128 mul1 = _mm_mul_ps(dot0, dot0);
+	__m128 sub0 = _mm_sub_ps(glm_one, mul0);
+	__m128 sub1 = _mm_sub_ps(glm_one, mul1);
+	__m128 mul2 = _mm_mul_ps(sub0, sub1);
+	
+	if(_mm_movemask_ps(_mm_cmplt_ss(mul2, glm_zero)) == 0)
+		return glm_zero;
+
+	__m128 sqt0 = _mm_sqrt_ps(mul2);
+	__m128 mul3 = _mm_mul_ps(eta, dot0);
+	__m128 add0 = _mm_add_ps(mul3, sqt0);
+	__m128 mul4 = _mm_mul_ps(add0, N);
+	__m128 mul5 = _mm_mul_ps(eta, I);
+	__m128 sub2 = _mm_sub_ps(mul5, mul4);
+
+	return sub2;
+}

glm/simd/matrix.h

@@ -0,0 +1,1034 @@
+#pragma once
+
+#include "geometric.h"
+
+static const __m128 GLM_VAR_USED _m128_rad_ps = _mm_set_ps1(3.141592653589793238462643383279f / 180.f);
+static const __m128 GLM_VAR_USED _m128_deg_ps = _mm_set_ps1(180.f / 3.141592653589793238462643383279f);
+
+template <typename matType>
+GLM_FUNC_QUALIFIER matType glm_comp_mul_ps
+(
+	__m128 const in1[4],
+	__m128 const in2[4],
+	__m128 out[4]
+)
+{
+	out[0] = _mm_mul_ps(in1[0], in2[0]);
+	out[1] = _mm_mul_ps(in1[1], in2[1]);
+	out[2] = _mm_mul_ps(in1[2], in2[2]);
+	out[3] = _mm_mul_ps(in1[3], in2[3]);
+}
+
+GLM_FUNC_QUALIFIER void glm_add_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
+{
+	{
+		out[0] = _mm_add_ps(in1[0], in2[0]);
+		out[1] = _mm_add_ps(in1[1], in2[1]);
+		out[2] = _mm_add_ps(in1[2], in2[2]);
+		out[3] = _mm_add_ps(in1[3], in2[3]);
+	}
+}
+
+GLM_FUNC_QUALIFIER void glm_sub_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
+{
+	{
+		out[0] = _mm_sub_ps(in1[0], in2[0]);
+		out[1] = _mm_sub_ps(in1[1], in2[1]);
+		out[2] = _mm_sub_ps(in1[2], in2[2]);
+		out[3] = _mm_sub_ps(in1[3], in2[3]);
+	}
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_mul_ps(__m128 const m[4], __m128 v)
+{
+	__m128 v0 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 v1 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 v2 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 v3 = _mm_shuffle_ps(v, v, _MM_SHUFFLE(3, 3, 3, 3));
+
+	__m128 m0 = _mm_mul_ps(m[0], v0);
+	__m128 m1 = _mm_mul_ps(m[1], v1);
+	__m128 m2 = _mm_mul_ps(m[2], v2);
+	__m128 m3 = _mm_mul_ps(m[3], v3);
+
+	__m128 a0 = _mm_add_ps(m0, m1);
+	__m128 a1 = _mm_add_ps(m2, m3);
+	__m128 a2 = _mm_add_ps(a0, a1);
+
+	return a2;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_mul_ps(__m128 v, __m128 const m[4])
+{
+	__m128 i0 = m[0];
+	__m128 i1 = m[1];
+	__m128 i2 = m[2];
+	__m128 i3 = m[3];
+
+	__m128 m0 = _mm_mul_ps(v, i0);
+	__m128 m1 = _mm_mul_ps(v, i1);
+	__m128 m2 = _mm_mul_ps(v, i2);
+	__m128 m3 = _mm_mul_ps(v, i3);
+
+	__m128 u0 = _mm_unpacklo_ps(m0, m1);
+	__m128 u1 = _mm_unpackhi_ps(m0, m1);
+	__m128 a0 = _mm_add_ps(u0, u1);
+
+	__m128 u2 = _mm_unpacklo_ps(m2, m3);
+	__m128 u3 = _mm_unpackhi_ps(m2, m3);
+	__m128 a1 = _mm_add_ps(u2, u3);
+
+	__m128 f0 = _mm_movelh_ps(a0, a1);
+	__m128 f1 = _mm_movehl_ps(a1, a0);
+	__m128 f2 = _mm_add_ps(f0, f1);
+
+	return f2;
+}
+
+GLM_FUNC_QUALIFIER void glm_mul_ps(__m128 const in1[4], __m128 const in2[4], __m128 out[4])
+{
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[0], in2[0], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[0] = a2;
+	}
+
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[1], in2[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[1] = a2;
+	}
+
+	{
+		__m128 e0 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[2], in2[2], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[2] = a2;
+	}
+
+	{
+		//(__m128&)_mm_shuffle_epi32(__m128i&)in2[0], _MM_SHUFFLE(3, 3, 3, 3))
+		__m128 e0 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 e1 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 e2 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 e3 = _mm_shuffle_ps(in2[3], in2[3], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 m0 = _mm_mul_ps(in1[0], e0);
+		__m128 m1 = _mm_mul_ps(in1[1], e1);
+		__m128 m2 = _mm_mul_ps(in1[2], e2);
+		__m128 m3 = _mm_mul_ps(in1[3], e3);
+
+		__m128 a0 = _mm_add_ps(m0, m1);
+		__m128 a1 = _mm_add_ps(m2, m3);
+		__m128 a2 = _mm_add_ps(a0, a1);
+
+		out[3] = a2;
+	}
+}
+
+GLM_FUNC_QUALIFIER void glm_transpose_ps(__m128 const in[4], __m128 out[4])
+{
+	__m128 tmp0 = _mm_shuffle_ps(in[0], in[1], 0x44);
+	__m128 tmp2 = _mm_shuffle_ps(in[0], in[1], 0xEE);
+	__m128 tmp1 = _mm_shuffle_ps(in[2], in[3], 0x44);
+	__m128 tmp3 = _mm_shuffle_ps(in[2], in[3], 0xEE);
+
+	out[0] = _mm_shuffle_ps(tmp0, tmp1, 0x88);
+	out[1] = _mm_shuffle_ps(tmp0, tmp1, 0xDD);
+	out[2] = _mm_shuffle_ps(tmp2, tmp3, 0x88);
+	out[3] = _mm_shuffle_ps(tmp2, tmp3, 0xDD);
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_slow_det_ps(__m128 const in[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0]
+	//						+ m[0][1] * Inverse[1][0]
+	//						+ m[0][2] * Inverse[2][0]
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_dot_ps(in[0], Row2);
+	return Det0;
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_detd_ps(__m128 const m[4])
+{
+	// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(
+
+	//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+	//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+	//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+	//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+	//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+	//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+
+	// First 2 columns
+ 	__m128 Swp2A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 1, 1, 2)));
+ 	__m128 Swp3A = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(3, 2, 3, 3)));
+	__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
+
+	// Second 2 columns
+	__m128 Swp2B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(3, 2, 3, 3)));
+	__m128 Swp3B = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(0, 1, 1, 2)));
+	__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
+
+	// Columns subtraction
+	__m128 SubE = _mm_sub_ps(MulA, MulB);
+
+	// Last 2 rows
+	__m128 Swp2C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[2]), _MM_SHUFFLE(0, 0, 1, 2)));
+	__m128 Swp3C = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[3]), _MM_SHUFFLE(1, 2, 0, 0)));
+	__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
+	__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
+
+	//tvec4<T, P> DetCof(
+	//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
+	//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
+	//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
+	//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
+
+	__m128 SubFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubE), _MM_SHUFFLE(2, 1, 0, 0)));
+	__m128 SwpFacA = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(0, 0, 0, 1)));
+	__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
+
+	__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
+	__m128 SubFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpB), _MM_SHUFFLE(3, 1, 1, 0)));//SubF[0], SubE[3], SubE[3], SubE[1];
+	__m128 SwpFacB = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(1, 1, 2, 2)));
+	__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
+
+	__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
+
+	__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
+	__m128 SubFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(SubTmpC), _MM_SHUFFLE(3, 3, 2, 0)));
+	__m128 SwpFacC = _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(m[1]), _MM_SHUFFLE(2, 3, 3, 3)));
+	__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
+
+	__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
+	__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
+
+	//return m[0][0] * DetCof[0]
+	//	 + m[0][1] * DetCof[1]
+	//	 + m[0][2] * DetCof[2]
+	//	 + m[0][3] * DetCof[3];
+
+	return glm_dot_ps(m[0], DetCof);
+}
+
+GLM_FUNC_QUALIFIER __m128 glm_det_ps(__m128 const m[4])
+{
+	// _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(add)
+
+	//T SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+	//T SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+	//T SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+	//T SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+	//T SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+	//T SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+
+	// First 2 columns
+ 	__m128 Swp2A = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 1, 1, 2));
+ 	__m128 Swp3A = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(3, 2, 3, 3));
+	__m128 MulA = _mm_mul_ps(Swp2A, Swp3A);
+
+	// Second 2 columns
+	__m128 Swp2B = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(3, 2, 3, 3));
+	__m128 Swp3B = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(0, 1, 1, 2));
+	__m128 MulB = _mm_mul_ps(Swp2B, Swp3B);
+
+	// Columns subtraction
+	__m128 SubE = _mm_sub_ps(MulA, MulB);
+
+	// Last 2 rows
+	__m128 Swp2C = _mm_shuffle_ps(m[2], m[2], _MM_SHUFFLE(0, 0, 1, 2));
+	__m128 Swp3C = _mm_shuffle_ps(m[3], m[3], _MM_SHUFFLE(1, 2, 0, 0));
+	__m128 MulC = _mm_mul_ps(Swp2C, Swp3C);
+	__m128 SubF = _mm_sub_ps(_mm_movehl_ps(MulC, MulC), MulC);
+
+	//tvec4<T, P> DetCof(
+	//	+ (m[1][1] * SubFactor00 - m[1][2] * SubFactor01 + m[1][3] * SubFactor02),
+	//	- (m[1][0] * SubFactor00 - m[1][2] * SubFactor03 + m[1][3] * SubFactor04),
+	//	+ (m[1][0] * SubFactor01 - m[1][1] * SubFactor03 + m[1][3] * SubFactor05),
+	//	- (m[1][0] * SubFactor02 - m[1][1] * SubFactor04 + m[1][2] * SubFactor05));
+
+	__m128 SubFacA = _mm_shuffle_ps(SubE, SubE, _MM_SHUFFLE(2, 1, 0, 0));
+	__m128 SwpFacA = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(0, 0, 0, 1));
+	__m128 MulFacA = _mm_mul_ps(SwpFacA, SubFacA);
+
+	__m128 SubTmpB = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(0, 0, 3, 1));
+	__m128 SubFacB = _mm_shuffle_ps(SubTmpB, SubTmpB, _MM_SHUFFLE(3, 1, 1, 0));//SubF[0], SubE[3], SubE[3], SubE[1];
+	__m128 SwpFacB = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(1, 1, 2, 2));
+	__m128 MulFacB = _mm_mul_ps(SwpFacB, SubFacB);
+
+	__m128 SubRes = _mm_sub_ps(MulFacA, MulFacB);
+
+	__m128 SubTmpC = _mm_shuffle_ps(SubE, SubF, _MM_SHUFFLE(1, 0, 2, 2));
+	__m128 SubFacC = _mm_shuffle_ps(SubTmpC, SubTmpC, _MM_SHUFFLE(3, 3, 2, 0));
+	__m128 SwpFacC = _mm_shuffle_ps(m[1], m[1], _MM_SHUFFLE(2, 3, 3, 3));
+	__m128 MulFacC = _mm_mul_ps(SwpFacC, SubFacC);
+
+	__m128 AddRes = _mm_add_ps(SubRes, MulFacC);
+	__m128 DetCof = _mm_mul_ps(AddRes, _mm_setr_ps( 1.0f,-1.0f, 1.0f,-1.0f));
+
+	//return m[0][0] * DetCof[0]
+	//	 + m[0][1] * DetCof[1]
+	//	 + m[0][2] * DetCof[2]
+	//	 + m[0][3] * DetCof[3];
+
+	return glm_dot_ps(m[0], DetCof);
+}
+
+GLM_FUNC_QUALIFIER void glm_inverse_ps(__m128 const in[4], __m128 out[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0] 
+	//						+ m[0][1] * Inverse[1][0] 
+	//						+ m[0][2] * Inverse[2][0] 
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_dot_ps(in[0], Row2);
+	__m128 Rcp0 = _mm_div_ps(glm_one, Det0);
+	//__m128 Rcp0 = _mm_rcp_ps(Det0);
+
+	//	Inverse /= Determinant;
+	out[0] = _mm_mul_ps(Inv0, Rcp0);
+	out[1] = _mm_mul_ps(Inv1, Rcp0);
+	out[2] = _mm_mul_ps(Inv2, Rcp0);
+	out[3] = _mm_mul_ps(Inv3, Rcp0);
+}
+
+GLM_FUNC_QUALIFIER void inverse_fast_ps(__m128 const in[4], __m128 out[4])
+{
+	__m128 Fac0;
+	{
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor00 = m[2][2] * m[3][3] - m[3][2] * m[2][3];
+		//	valType SubFactor06 = m[1][2] * m[3][3] - m[3][2] * m[1][3];
+		//	valType SubFactor13 = m[1][2] * m[2][3] - m[2][2] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac0 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac1;
+	{
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor01 = m[2][1] * m[3][3] - m[3][1] * m[2][3];
+		//	valType SubFactor07 = m[1][1] * m[3][3] - m[3][1] * m[1][3];
+		//	valType SubFactor14 = m[1][1] * m[2][3] - m[2][1] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac1 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+
+	__m128 Fac2;
+	{
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor02 = m[2][1] * m[3][2] - m[3][1] * m[2][2];
+		//	valType SubFactor08 = m[1][1] * m[3][2] - m[3][1] * m[1][2];
+		//	valType SubFactor15 = m[1][1] * m[2][2] - m[2][1] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac2 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac3;
+	{
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor03 = m[2][0] * m[3][3] - m[3][0] * m[2][3];
+		//	valType SubFactor09 = m[1][0] * m[3][3] - m[3][0] * m[1][3];
+		//	valType SubFactor16 = m[1][0] * m[2][3] - m[2][0] * m[1][3];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(3, 3, 3, 3));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(3, 3, 3, 3));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac3 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac4;
+	{
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor04 = m[2][0] * m[3][2] - m[3][0] * m[2][2];
+		//	valType SubFactor10 = m[1][0] * m[3][2] - m[3][0] * m[1][2];
+		//	valType SubFactor17 = m[1][0] * m[2][2] - m[2][0] * m[1][2];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(2, 2, 2, 2));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(2, 2, 2, 2));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac4 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 Fac5;
+	{
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor05 = m[2][0] * m[3][1] - m[3][0] * m[2][1];
+		//	valType SubFactor12 = m[1][0] * m[3][1] - m[3][0] * m[1][1];
+		//	valType SubFactor18 = m[1][0] * m[2][1] - m[2][0] * m[1][1];
+
+		__m128 Swp0a = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(1, 1, 1, 1));
+		__m128 Swp0b = _mm_shuffle_ps(in[3], in[2], _MM_SHUFFLE(0, 0, 0, 0));
+
+		__m128 Swp00 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(0, 0, 0, 0));
+		__m128 Swp01 = _mm_shuffle_ps(Swp0a, Swp0a, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp02 = _mm_shuffle_ps(Swp0b, Swp0b, _MM_SHUFFLE(2, 0, 0, 0));
+		__m128 Swp03 = _mm_shuffle_ps(in[2], in[1], _MM_SHUFFLE(1, 1, 1, 1));
+
+		__m128 Mul00 = _mm_mul_ps(Swp00, Swp01);
+		__m128 Mul01 = _mm_mul_ps(Swp02, Swp03);
+		Fac5 = _mm_sub_ps(Mul00, Mul01);
+	}
+
+	__m128 SignA = _mm_set_ps( 1.0f,-1.0f, 1.0f,-1.0f);
+	__m128 SignB = _mm_set_ps(-1.0f, 1.0f,-1.0f, 1.0f);
+
+	// m[1][0]
+	// m[0][0]
+	// m[0][0]
+	// m[0][0]
+	__m128 Temp0 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Vec0 = _mm_shuffle_ps(Temp0, Temp0, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][1]
+	// m[0][1]
+	// m[0][1]
+	// m[0][1]
+	__m128 Temp1 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 Vec1 = _mm_shuffle_ps(Temp1, Temp1, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][2]
+	// m[0][2]
+	// m[0][2]
+	// m[0][2]
+	__m128 Temp2 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 Vec2 = _mm_shuffle_ps(Temp2, Temp2, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// m[1][3]
+	// m[0][3]
+	// m[0][3]
+	// m[0][3]
+	__m128 Temp3 = _mm_shuffle_ps(in[1], in[0], _MM_SHUFFLE(3, 3, 3, 3));
+	__m128 Vec3 = _mm_shuffle_ps(Temp3, Temp3, _MM_SHUFFLE(2, 2, 2, 0));
+
+	// col0
+	// + (Vec1[0] * Fac0[0] - Vec2[0] * Fac1[0] + Vec3[0] * Fac2[0]),
+	// - (Vec1[1] * Fac0[1] - Vec2[1] * Fac1[1] + Vec3[1] * Fac2[1]),
+	// + (Vec1[2] * Fac0[2] - Vec2[2] * Fac1[2] + Vec3[2] * Fac2[2]),
+	// - (Vec1[3] * Fac0[3] - Vec2[3] * Fac1[3] + Vec3[3] * Fac2[3]),
+	__m128 Mul00 = _mm_mul_ps(Vec1, Fac0);
+	__m128 Mul01 = _mm_mul_ps(Vec2, Fac1);
+	__m128 Mul02 = _mm_mul_ps(Vec3, Fac2);
+	__m128 Sub00 = _mm_sub_ps(Mul00, Mul01);
+	__m128 Add00 = _mm_add_ps(Sub00, Mul02);
+	__m128 Inv0 = _mm_mul_ps(SignB, Add00);
+
+	// col1
+	// - (Vec0[0] * Fac0[0] - Vec2[0] * Fac3[0] + Vec3[0] * Fac4[0]),
+	// + (Vec0[0] * Fac0[1] - Vec2[1] * Fac3[1] + Vec3[1] * Fac4[1]),
+	// - (Vec0[0] * Fac0[2] - Vec2[2] * Fac3[2] + Vec3[2] * Fac4[2]),
+	// + (Vec0[0] * Fac0[3] - Vec2[3] * Fac3[3] + Vec3[3] * Fac4[3]),
+	__m128 Mul03 = _mm_mul_ps(Vec0, Fac0);
+	__m128 Mul04 = _mm_mul_ps(Vec2, Fac3);
+	__m128 Mul05 = _mm_mul_ps(Vec3, Fac4);
+	__m128 Sub01 = _mm_sub_ps(Mul03, Mul04);
+	__m128 Add01 = _mm_add_ps(Sub01, Mul05);
+	__m128 Inv1 = _mm_mul_ps(SignA, Add01);
+
+	// col2
+	// + (Vec0[0] * Fac1[0] - Vec1[0] * Fac3[0] + Vec3[0] * Fac5[0]),
+	// - (Vec0[0] * Fac1[1] - Vec1[1] * Fac3[1] + Vec3[1] * Fac5[1]),
+	// + (Vec0[0] * Fac1[2] - Vec1[2] * Fac3[2] + Vec3[2] * Fac5[2]),
+	// - (Vec0[0] * Fac1[3] - Vec1[3] * Fac3[3] + Vec3[3] * Fac5[3]),
+	__m128 Mul06 = _mm_mul_ps(Vec0, Fac1);
+	__m128 Mul07 = _mm_mul_ps(Vec1, Fac3);
+	__m128 Mul08 = _mm_mul_ps(Vec3, Fac5);
+	__m128 Sub02 = _mm_sub_ps(Mul06, Mul07);
+	__m128 Add02 = _mm_add_ps(Sub02, Mul08);
+	__m128 Inv2 = _mm_mul_ps(SignB, Add02);
+
+	// col3
+	// - (Vec1[0] * Fac2[0] - Vec1[0] * Fac4[0] + Vec2[0] * Fac5[0]),
+	// + (Vec1[0] * Fac2[1] - Vec1[1] * Fac4[1] + Vec2[1] * Fac5[1]),
+	// - (Vec1[0] * Fac2[2] - Vec1[2] * Fac4[2] + Vec2[2] * Fac5[2]),
+	// + (Vec1[0] * Fac2[3] - Vec1[3] * Fac4[3] + Vec2[3] * Fac5[3]));
+	__m128 Mul09 = _mm_mul_ps(Vec0, Fac2);
+	__m128 Mul10 = _mm_mul_ps(Vec1, Fac4);
+	__m128 Mul11 = _mm_mul_ps(Vec2, Fac5);
+	__m128 Sub03 = _mm_sub_ps(Mul09, Mul10);
+	__m128 Add03 = _mm_add_ps(Sub03, Mul11);
+	__m128 Inv3 = _mm_mul_ps(SignA, Add03);
+
+	__m128 Row0 = _mm_shuffle_ps(Inv0, Inv1, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row1 = _mm_shuffle_ps(Inv2, Inv3, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Row2 = _mm_shuffle_ps(Row0, Row1, _MM_SHUFFLE(2, 0, 2, 0));
+
+	//	valType Determinant = m[0][0] * Inverse[0][0] 
+	//						+ m[0][1] * Inverse[1][0] 
+	//						+ m[0][2] * Inverse[2][0] 
+	//						+ m[0][3] * Inverse[3][0];
+	__m128 Det0 = glm_dot_ps(in[0], Row2);
+	__m128 Rcp0 = _mm_rcp_ps(Det0);
+	//__m128 Rcp0 = _mm_div_ps(one, Det0);
+	//	Inverse /= Determinant;
+	out[0] = _mm_mul_ps(Inv0, Rcp0);
+	out[1] = _mm_mul_ps(Inv1, Rcp0);
+	out[2] = _mm_mul_ps(Inv2, Rcp0);
+	out[3] = _mm_mul_ps(Inv3, Rcp0);
+}
+/*
+GLM_FUNC_QUALIFIER void glm_rotate_ps(__m128 const in[4], float Angle, float const v[3], __m128 out[4])
+{
+	float a = glm::radians(Angle);
+	float c = cos(a);
+	float s = sin(a);
+
+	glm::vec4 AxisA(v[0], v[1], v[2], float(0));
+	__m128 AxisB = _mm_set_ps(AxisA.w, AxisA.z, AxisA.y, AxisA.x);
+	__m128 AxisC = detail::sse_nrm_ps(AxisB);
+
+	__m128 Cos0 = _mm_set_ss(c);
+	__m128 CosA = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 Sin0 = _mm_set_ss(s);
+	__m128 SinA = _mm_shuffle_ps(Sin0, Sin0, _MM_SHUFFLE(0, 0, 0, 0));
+
+	// tvec3<T, P> temp = (valType(1) - c) * axis;
+	__m128 Temp0 = _mm_sub_ps(one, CosA);
+	__m128 Temp1 = _mm_mul_ps(Temp0, AxisC);
+	
+	//Rotate[0][0] = c + temp[0] * axis[0];
+	//Rotate[0][1] = 0 + temp[0] * axis[1] + s * axis[2];
+	//Rotate[0][2] = 0 + temp[0] * axis[2] - s * axis[1];
+	__m128 Axis0 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(0, 0, 0, 0));
+	__m128 TmpA0 = _mm_mul_ps(Axis0, AxisC);
+	__m128 CosA0 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 1, 0));
+	__m128 TmpA1 = _mm_add_ps(CosA0, TmpA0);
+	__m128 SinA0 = SinA;//_mm_set_ps(0.0f, s, -s, 0.0f);
+	__m128 TmpA2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 1, 2, 3));
+	__m128 TmpA3 = _mm_mul_ps(SinA0, TmpA2);
+	__m128 TmpA4 = _mm_add_ps(TmpA1, TmpA3);
+
+	//Rotate[1][0] = 0 + temp[1] * axis[0] - s * axis[2];
+	//Rotate[1][1] = c + temp[1] * axis[1];
+	//Rotate[1][2] = 0 + temp[1] * axis[2] + s * axis[0];
+	__m128 Axis1 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(1, 1, 1, 1));
+	__m128 TmpB0 = _mm_mul_ps(Axis1, AxisC);
+	__m128 CosA1 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 1, 0, 1));
+	__m128 TmpB1 = _mm_add_ps(CosA1, TmpB0);
+	__m128 SinB0 = SinA;//_mm_set_ps(-s, 0.0f, s, 0.0f);
+	__m128 TmpB2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 0, 3, 2));
+	__m128 TmpB3 = _mm_mul_ps(SinA0, TmpB2);
+	__m128 TmpB4 = _mm_add_ps(TmpB1, TmpB3);
+
+	//Rotate[2][0] = 0 + temp[2] * axis[0] + s * axis[1];
+	//Rotate[2][1] = 0 + temp[2] * axis[1] - s * axis[0];
+	//Rotate[2][2] = c + temp[2] * axis[2];
+	__m128 Axis2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(2, 2, 2, 2));
+	__m128 TmpC0 = _mm_mul_ps(Axis2, AxisC);
+	__m128 CosA2 = _mm_shuffle_ps(Cos0, Cos0, _MM_SHUFFLE(1, 0, 1, 1));
+	__m128 TmpC1 = _mm_add_ps(CosA2, TmpC0);
+	__m128 SinC0 = SinA;//_mm_set_ps(s, -s, 0.0f, 0.0f);
+	__m128 TmpC2 = _mm_shuffle_ps(AxisC, AxisC, _MM_SHUFFLE(3, 3, 0, 1));
+	__m128 TmpC3 = _mm_mul_ps(SinA0, TmpC2);
+	__m128 TmpC4 = _mm_add_ps(TmpC1, TmpC3);
+
+	__m128 Result[4];
+	Result[0] = TmpA4;
+	Result[1] = TmpB4;
+	Result[2] = TmpC4;
+	Result[3] = _mm_set_ps(1, 0, 0, 0);
+
+	//tmat4x4<valType> Result(uninitialize);
+	//Result[0] = m[0] * Rotate[0][0] + m[1] * Rotate[0][1] + m[2] * Rotate[0][2];
+	//Result[1] = m[0] * Rotate[1][0] + m[1] * Rotate[1][1] + m[2] * Rotate[1][2];
+	//Result[2] = m[0] * Rotate[2][0] + m[1] * Rotate[2][1] + m[2] * Rotate[2][2];
+	//Result[3] = m[3];
+	//return Result;
+	sse_mul_ps(in, Result, out);
+}
+*/
+GLM_FUNC_QUALIFIER void glm_outer_ps(__m128 const & c, __m128 const & r, __m128 out[4])
+{
+	out[0] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(0, 0, 0, 0)));
+	out[1] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(1, 1, 1, 1)));
+	out[2] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(2, 2, 2, 2)));
+	out[3] = _mm_mul_ps(c, _mm_shuffle_ps(r, r, _MM_SHUFFLE(3, 3, 3, 3)));
+}

glm/simd/platform.h

@@ -0,0 +1,280 @@
+/// @ref simd
+/// @file glm/simd/platform.h
+
+#pragma once
+
+///////////////////////////////////////////////////////////////////////////////////
+// Platform
+
+#define GLM_PLATFORM_UNKNOWN		0x00000000
+#define GLM_PLATFORM_WINDOWS		0x00010000
+#define GLM_PLATFORM_LINUX			0x00020000
+#define GLM_PLATFORM_APPLE			0x00040000
+//#define GLM_PLATFORM_IOS			0x00080000
+#define GLM_PLATFORM_ANDROID		0x00100000
+#define GLM_PLATFORM_CHROME_NACL	0x00200000
+#define GLM_PLATFORM_UNIX			0x00400000
+#define GLM_PLATFORM_QNXNTO			0x00800000
+#define GLM_PLATFORM_WINCE			0x01000000
+#define GLM_PLATFORM_CYGWIN			0x02000000
+
+#ifdef GLM_FORCE_PLATFORM_UNKNOWN
+#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
+#elif defined(__CYGWIN__)
+#	define GLM_PLATFORM GLM_PLATFORM_CYGWIN
+#elif defined(__QNXNTO__)
+#	define GLM_PLATFORM GLM_PLATFORM_QNXNTO
+#elif defined(__APPLE__)
+#	define GLM_PLATFORM GLM_PLATFORM_APPLE
+#elif defined(WINCE)
+#	define GLM_PLATFORM GLM_PLATFORM_WINCE
+#elif defined(_WIN32)
+#	define GLM_PLATFORM GLM_PLATFORM_WINDOWS
+#elif defined(__native_client__)
+#	define GLM_PLATFORM GLM_PLATFORM_CHROME_NACL
+#elif defined(__ANDROID__)
+#	define GLM_PLATFORM GLM_PLATFORM_ANDROID
+#elif defined(__linux)
+#	define GLM_PLATFORM GLM_PLATFORM_LINUX
+#elif defined(__unix)
+#	define GLM_PLATFORM GLM_PLATFORM_UNIX
+#else
+#	define GLM_PLATFORM GLM_PLATFORM_UNKNOWN
+#endif//
+
+// Report platform detection
+#if(defined(GLM_MESSAGES) && !defined(GLM_MESSAGE_PLATFORM_DISPLAYED))
+#	define GLM_MESSAGE_PLATFORM_DISPLAYED
+#	if(GLM_PLATFORM & GLM_PLATFORM_QNXNTO)
+#		pragma message("GLM: QNX platform detected")
+//#	elif(GLM_PLATFORM & GLM_PLATFORM_IOS)
+//#		pragma message("GLM: iOS platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_APPLE)
+#		pragma message("GLM: Apple platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_WINCE)
+#		pragma message("GLM: WinCE platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_WINDOWS)
+#		pragma message("GLM: Windows platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_CHROME_NACL)
+#		pragma message("GLM: Native Client detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_ANDROID)
+#		pragma message("GLM: Android platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_LINUX)
+#		pragma message("GLM: Linux platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_UNIX)
+#		pragma message("GLM: UNIX platform detected")
+#	elif(GLM_PLATFORM & GLM_PLATFORM_UNKNOWN)
+#		pragma message("GLM: platform unknown")
+#	else
+#		pragma message("GLM: platform not detected")
+#	endif
+#endif//GLM_MESSAGE
+
+///////////////////////////////////////////////////////////////////////////////////
+// Compiler
+
+#define GLM_COMPILER_UNKNOWN		0x00000000
+
+// Intel
+#define GLM_COMPILER_INTEL			0x00100000
+#define GLM_COMPILER_INTEL12		0x00100010
+#define GLM_COMPILER_INTEL12_1		0x00100020
+#define GLM_COMPILER_INTEL13		0x00100030
+#define GLM_COMPILER_INTEL14		0x00100040
+#define GLM_COMPILER_INTEL15		0x00100050
+#define GLM_COMPILER_INTEL16		0x00100060
+
+// Visual C++ defines
+#define GLM_COMPILER_VC				0x01000000
+#define GLM_COMPILER_VC2010			0x01000090
+#define GLM_COMPILER_VC2012			0x010000A0
+#define GLM_COMPILER_VC2013			0x010000B0
+#define GLM_COMPILER_VC2015			0x010000C0
+
+// GCC defines
+#define GLM_COMPILER_GCC			0x02000000
+#define GLM_COMPILER_GCC44			0x020000B0
+#define GLM_COMPILER_GCC45			0x020000C0
+#define GLM_COMPILER_GCC46			0x020000D0
+#define GLM_COMPILER_GCC47			0x020000E0
+#define GLM_COMPILER_GCC48			0x020000F0
+#define GLM_COMPILER_GCC49			0x02000100
+#define GLM_COMPILER_GCC50			0x02000200
+#define GLM_COMPILER_GCC51			0x02000300
+#define GLM_COMPILER_GCC52			0x02000400
+#define GLM_COMPILER_GCC53			0x02000500
+#define GLM_COMPILER_GCC60			0x02000600
+
+// CUDA
+#define GLM_COMPILER_CUDA			0x10000000
+#define GLM_COMPILER_CUDA40			0x10000040
+#define GLM_COMPILER_CUDA41			0x10000050
+#define GLM_COMPILER_CUDA42			0x10000060
+#define GLM_COMPILER_CUDA50			0x10000070
+#define GLM_COMPILER_CUDA60			0x10000080
+#define GLM_COMPILER_CUDA65			0x10000090
+#define GLM_COMPILER_CUDA70			0x100000A0
+#define GLM_COMPILER_CUDA75			0x100000B0
+
+// LLVM
+#define GLM_COMPILER_LLVM			0x20000000
+#define GLM_COMPILER_LLVM32			0x20000030
+#define GLM_COMPILER_LLVM33			0x20000040
+#define GLM_COMPILER_LLVM34			0x20000050
+#define GLM_COMPILER_LLVM35			0x20000060
+#define GLM_COMPILER_LLVM36			0x20000070
+#define GLM_COMPILER_LLVM37			0x20000080
+#define GLM_COMPILER_LLVM38			0x20000090
+#define GLM_COMPILER_LLVM39			0x200000A0
+
+// Apple Clang
+#define GLM_COMPILER_APPLE_CLANG	0x40000000
+#define GLM_COMPILER_APPLE_CLANG40	0x40000010
+#define GLM_COMPILER_APPLE_CLANG41	0x40000020
+#define GLM_COMPILER_APPLE_CLANG42	0x40000030
+#define GLM_COMPILER_APPLE_CLANG50	0x40000040
+#define GLM_COMPILER_APPLE_CLANG51	0x40000050
+#define GLM_COMPILER_APPLE_CLANG60	0x40000060
+#define GLM_COMPILER_APPLE_CLANG61	0x40000070
+
+// Build model
+#define GLM_MODEL_32				0x00000010
+#define GLM_MODEL_64				0x00000020
+
+// Force generic C++ compiler
+#ifdef GLM_FORCE_COMPILER_UNKNOWN
+#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
+
+#elif defined(__INTEL_COMPILER)
+#	if __INTEL_COMPILER == 1200
+#		define GLM_COMPILER GLM_COMPILER_INTEL12
+#	elif __INTEL_COMPILER == 1210
+#		define GLM_COMPILER GLM_COMPILER_INTEL12_1
+#	elif __INTEL_COMPILER == 1300
+#		define GLM_COMPILER GLM_COMPILER_INTEL13
+#	elif __INTEL_COMPILER == 1400
+#		define GLM_COMPILER GLM_COMPILER_INTEL14
+#	elif __INTEL_COMPILER == 1500
+#		define GLM_COMPILER GLM_COMPILER_INTEL15
+#	elif __INTEL_COMPILER >= 1600
+#		define GLM_COMPILER GLM_COMPILER_INTEL16
+#	else
+#		define GLM_COMPILER GLM_COMPILER_INTEL
+#	endif
+
+// CUDA
+#elif defined(__CUDACC__)
+#	if !defined(CUDA_VERSION) && !defined(GLM_FORCE_CUDA)
+#		include <cuda.h>  // make sure version is defined since nvcc does not define it itself!
+#	endif
+#	if CUDA_VERSION < 3000
+#		error "GLM requires CUDA 3.0 or higher"
+#	else
+#		define GLM_COMPILER GLM_COMPILER_CUDA
+#	endif
+
+// Visual C++
+#elif defined(_MSC_VER)
+#	if _MSC_VER < 1600
+#		error "GLM requires Visual C++ 2010 or higher"
+#	elif _MSC_VER == 1600
+#		define GLM_COMPILER GLM_COMPILER_VC2010
+#	elif _MSC_VER == 1700
+#		define GLM_COMPILER GLM_COMPILER_VC2012
+#	elif _MSC_VER == 1800
+#		define GLM_COMPILER GLM_COMPILER_VC2013
+#	elif _MSC_VER >= 1900
+#		define GLM_COMPILER GLM_COMPILER_VC2015
+#	else//_MSC_VER
+#		define GLM_COMPILER GLM_COMPILER_VC
+#	endif//_MSC_VER
+
+// Clang
+#elif defined(__clang__)
+#	if GLM_PLATFORM & GLM_PLATFORM_APPLE
+#		if __clang_major__ == 4 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG40
+#		elif __clang_major__ == 4 && __clang_minor__ == 1
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG41
+#		elif __clang_major__ == 4 && __clang_minor__ == 2
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG42
+#		elif __clang_major__ == 5 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG50
+#		elif __clang_major__ == 5 && __clang_minor__ == 1
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG51
+#		elif __clang_major__ == 6 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG60
+#		elif __clang_major__ == 6 && __clang_minor__ >= 1
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG61
+#		elif __clang_major__ >= 7
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG61
+#		else
+#			define GLM_COMPILER GLM_COMPILER_APPLE_CLANG
+#		endif
+#	else
+#		if __clang_major__ == 3 && __clang_minor__ == 0
+#			define GLM_COMPILER GLM_COMPILER_LLVM30
+#		elif __clang_major__ == 3 && __clang_minor__ == 1
+#			define GLM_COMPILER GLM_COMPILER_LLVM31
+#		elif __clang_major__ == 3 && __clang_minor__ == 2
+#			define GLM_COMPILER GLM_COMPILER_LLVM32
+#		elif __clang_major__ == 3 && __clang_minor__ == 3
+#			define GLM_COMPILER GLM_COMPILER_LLVM33
+#		elif __clang_major__ == 3 && __clang_minor__ == 4
+#			define GLM_COMPILER GLM_COMPILER_LLVM34
+#		elif __clang_major__ == 3 && __clang_minor__ == 5
+#			define GLM_COMPILER GLM_COMPILER_LLVM35
+#		elif __clang_major__ == 3 && __clang_minor__ == 6
+#			define GLM_COMPILER GLM_COMPILER_LLVM36
+#		elif __clang_major__ == 3 && __clang_minor__ == 7
+#			define GLM_COMPILER GLM_COMPILER_LLVM37
+#		elif __clang_major__ == 3 && __clang_minor__ == 8
+#			define GLM_COMPILER GLM_COMPILER_LLVM38
+#		elif __clang_major__ == 3 && __clang_minor__ >= 9
+#			define GLM_COMPILER GLM_COMPILER_LLVM39
+#		elif __clang_major__ >= 4
+#			define GLM_COMPILER GLM_COMPILER_LLVM39
+#		else
+#			define GLM_COMPILER GLM_COMPILER_LLVM
+#		endif
+#	endif
+
+// G++
+#elif defined(__GNUC__) || defined(__MINGW32__)
+#	if (__GNUC__ == 4) && (__GNUC_MINOR__ == 2)
+#		define GLM_COMPILER (GLM_COMPILER_GCC42)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 3)
+#		define GLM_COMPILER (GLM_COMPILER_GCC43)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 4)
+#		define GLM_COMPILER (GLM_COMPILER_GCC44)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 5)
+#		define GLM_COMPILER (GLM_COMPILER_GCC45)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 6)
+#		define GLM_COMPILER (GLM_COMPILER_GCC46)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 7)
+#		define GLM_COMPILER (GLM_COMPILER_GCC47)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ == 8)
+#		define GLM_COMPILER (GLM_COMPILER_GCC48)
+#	elif (__GNUC__ == 4) && (__GNUC_MINOR__ >= 9)
+#		define GLM_COMPILER (GLM_COMPILER_GCC49)
+#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 0)
+#		define GLM_COMPILER (GLM_COMPILER_GCC50)
+#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 1)
+#		define GLM_COMPILER (GLM_COMPILER_GCC51)
+#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ == 2)
+#		define GLM_COMPILER (GLM_COMPILER_GCC52)
+#	elif (__GNUC__ == 5) && (__GNUC_MINOR__ >= 3)
+#		define GLM_COMPILER (GLM_COMPILER_GCC53)
+#	elif (__GNUC__ >= 6)
+#		define GLM_COMPILER (GLM_COMPILER_GCC60)
+#	else
+#		define GLM_COMPILER (GLM_COMPILER_GCC)
+#	endif
+
+#else
+#	define GLM_COMPILER GLM_COMPILER_UNKNOWN
+#endif
+
+#ifndef GLM_COMPILER
+#	error "GLM_COMPILER undefined, your compiler may not be supported by GLM. Add #define GLM_COMPILER 0 to ignore this message."
+#endif//GLM_COMPILER

test/core/core_type_vec1.cpp

@@ -96,7 +96,7 @@ int test_vec1_ctor()
 	}
 #endif
 */
-#if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 	{
 		glm::vec2 A = glm::vec2(1.0f, 2.0f);
 		glm::vec2 B = A.xy;
@@ -107,7 +107,7 @@ int test_vec1_ctor()
 		Error += glm::all(glm::equal(A, C)) ? 0 : 1;
 		Error += glm::all(glm::equal(A, D)) ? 0 : 1;
 	}
-#endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#endif//GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 	{
 		glm::vec2 A = glm::vec2(2.0f);

test/core/core_type_vec2.cpp

@@ -262,7 +262,7 @@ int test_vec2_ctor()
 	}
 #endif
 
-#if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 	{
 		glm::vec2 A = glm::vec2(1.0f, 2.0f);
 		glm::vec2 B = A.xy;
@@ -273,7 +273,7 @@ int test_vec2_ctor()
 		Error += glm::all(glm::equal(A, C)) ? 0 : 1;
 		Error += glm::all(glm::equal(A, D)) ? 0 : 1;
 	}
-#endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#endif//GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 	{
 		glm::vec2 A = glm::vec2(2.0f);

test/core/core_type_vec3.cpp

@@ -71,7 +71,7 @@ int test_vec3_ctor()
 	}
 #endif
 
-#if(GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE))
+#if(GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE))
 	{
 		glm::vec3 A = glm::vec3(1.0f, 2.0f, 3.0f);
 		glm::vec3 B = A.xyz;
@@ -90,7 +90,7 @@ int test_vec3_ctor()
 		Error += glm::all(glm::equal(A, G)) ? 0 : 1;
 		Error += glm::all(glm::equal(A, H)) ? 0 : 1;
 	}
-#endif//(GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE))
+#endif//(GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE))
 
 	{
 		glm::vec3 A(1);

test/core/core_type_vec4.cpp

@@ -98,7 +98,7 @@ int test_vec4_ctor()
 	}
 #endif
 
-#if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 	{
 		glm::vec4 A = glm::vec4(1.0f, 2.0f, 3.0f, 4.0f);
 		glm::vec4 B = A.xyzw;
@@ -127,7 +127,7 @@ int test_vec4_ctor()
 		Error += glm::all(glm::equal(A, L)) ? 0 : 1;
 		Error += glm::all(glm::equal(A, M)) ? 0 : 1;
 	}
-#endif// GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE)
+#endif// GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE)
 
 	{
 		glm::vec4 A(1);
@@ -334,7 +334,7 @@ int test_vec4_swizzle_partial()
 
 	glm::vec4 A(1, 2, 3, 4);
 
-#	if GLM_HAS_ANONYMOUS_UNION && defined(GLM_SWIZZLE_RELAX)
+#	if GLM_HAS_UNRESTRICTED_UNIONS && defined(GLM_SWIZZLE_RELAX)
 	{
 		glm::vec4 B(A.xy, A.zw);
 		Error += A == B ? 0 : 1;