Merge pull request #25 from mackron/SIMD

SIMD Improvements and Additions

glm/core/intrinsic_matrix.inl

@@ -46,7 +46,7 @@ GLM_FUNC_QUALIFIER matType sse_comp_mul_ps
 	out[3] = _mm_mul_ps(in1[3], in2[3]);
 }
 
-GLM_FUNC_QUALIFIER void sse_add_ps(__m128 in1[4], __m128 in2[4], __m128 out[4])
+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]);
@@ -56,7 +56,7 @@ GLM_FUNC_QUALIFIER void sse_add_ps(__m128 in1[4], __m128 in2[4], __m128 out[4])
 	}
 }
 
-GLM_FUNC_QUALIFIER void sse_sub_ps(__m128 in1[4], __m128 in2[4], __m128 out[4])
+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]);
@@ -66,7 +66,7 @@ GLM_FUNC_QUALIFIER void sse_sub_ps(__m128 in1[4], __m128 in2[4], __m128 out[4])
 	}
 }
 
-GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 m[4], __m128 v)
+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));
@@ -85,7 +85,7 @@ GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 m[4], __m128 v)
 	return a2;
 }
 
-GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 v, __m128 m[4])
+GLM_FUNC_QUALIFIER __m128 sse_mul_ps(__m128 v, __m128 const m[4])
 {
 	__m128 i0 = m[0];
 	__m128 i1 = m[1];

glm/gtx/simd_mat4.hpp

@@ -91,6 +91,8 @@ namespace detail
 			fvec4SIMD const & v3);
 		explicit fmat4x4SIMD(
 			tmat4x4<float> const & m);
+        explicit fmat4x4SIMD(
+            __m128 const in[4]);
 
 		// Conversions
 		//template <typename U> 

glm/gtx/simd_mat4.inl

@@ -26,7 +26,14 @@ GLM_FUNC_QUALIFIER fmat4x4SIMD::size_type fmat4x4SIMD::row_size()
 }
 
 GLM_FUNC_QUALIFIER fmat4x4SIMD::fmat4x4SIMD()
-{}
+{
+#ifndef GLM_SIMD_ENABLE_DEFAULT_INIT
+    this->Data[0] = fvec4SIMD(1.0f, 0, 0, 0);
+	this->Data[1] = fvec4SIMD(0, 1.0f, 0, 0);
+	this->Data[2] = fvec4SIMD(0, 0, 1.0f, 0);
+	this->Data[3] = fvec4SIMD(0, 0, 0, 1.0f);
+#endif
+}
 
 GLM_FUNC_QUALIFIER fmat4x4SIMD::fmat4x4SIMD(float const & s)
 {
@@ -75,6 +82,17 @@ GLM_FUNC_QUALIFIER fmat4x4SIMD::fmat4x4SIMD
 	this->Data[3] = fvec4SIMD(m[3]);
 }
 
+GLM_FUNC_QUALIFIER fmat4x4SIMD::fmat4x4SIMD
+(
+    __m128 const in[4]
+)
+{
+    this->Data[0] = in[0];
+    this->Data[1] = in[1];
+    this->Data[2] = in[2];
+    this->Data[3] = in[3];
+}
+
 //////////////////////////////////////
 // Accesses
 
@@ -232,6 +250,274 @@ GLM_FUNC_QUALIFIER fmat4x4SIMD & fmat4x4SIMD::operator-- ()
     return *this;
 }
 
+
+//////////////////////////////////////////////////////////////
+// Binary operators
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator+
+(
+    const fmat4x4SIMD &m,
+    float const & s
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] + s,
+        m[1] + s,
+        m[2] + s,
+        m[3] + s
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator+
+(
+    float const & s,
+    const fmat4x4SIMD &m
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] + s,
+        m[1] + s,
+        m[2] + s,
+        m[3] + s
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator+
+(
+    const fmat4x4SIMD &m1,
+    const fmat4x4SIMD &m2
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m1[0] + m2[0],
+        m1[1] + m2[1],
+        m1[2] + m2[2],
+        m1[3] + m2[3]
+    );
+}
+
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator-
+(
+    const fmat4x4SIMD &m,
+    float const & s
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] - s,
+        m[1] - s,
+        m[2] - s,
+        m[3] - s
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator-
+(
+    float const & s,
+    const fmat4x4SIMD &m
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        s - m[0],
+        s - m[1],
+        s - m[2],
+        s - m[3]
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator-
+(
+    const fmat4x4SIMD &m1,
+    const fmat4x4SIMD &m2
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m1[0] - m2[0],
+        m1[1] - m2[1],
+        m1[2] - m2[2],
+        m1[3] - m2[3]
+    );
+}
+
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator*
+(
+    const fmat4x4SIMD &m,
+    float const & s
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] * s,
+        m[1] * s,
+        m[2] * s,
+        m[3] * s
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator*
+(
+    float const & s,
+    const fmat4x4SIMD &m
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] * s,
+        m[1] * s,
+        m[2] * s,
+        m[3] * s
+    );
+}
+
+GLM_FUNC_QUALIFIER fvec4SIMD operator*
+(
+    const fmat4x4SIMD &m,
+    fvec4SIMD const & v
+)
+{
+    return sse_mul_ps(&m.Data[0].Data, v.Data);
+}
+
+GLM_FUNC_QUALIFIER fvec4SIMD operator*
+(
+    fvec4SIMD const & v,
+    const fmat4x4SIMD &m
+)
+{
+    return sse_mul_ps(v.Data, &m.Data[0].Data);
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator*
+(
+    const fmat4x4SIMD &m1,
+    const fmat4x4SIMD &m2
+)
+{
+    fmat4x4SIMD result;
+    sse_mul_ps(&m1.Data[0].Data, &m2.Data[0].Data, &result.Data[0].Data);
+    
+    return result;
+}
+    
+
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator/
+(
+    const fmat4x4SIMD &m,
+    float const & s
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] / s,
+        m[1] / s,
+        m[2] / s,
+        m[3] / s
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator/
+(
+    float const & s,
+    const fmat4x4SIMD &m
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        s / m[0],
+        s / m[1],
+        s / m[2],
+        s / m[3]
+    );
+}
+
+GLM_FUNC_QUALIFIER fvec4SIMD operator/
+(
+    const fmat4x4SIMD &m,
+    fvec4SIMD const & v
+)
+{
+    return inverse(m) * v;
+}
+
+GLM_FUNC_QUALIFIER fvec4SIMD operator/
+(
+    fvec4SIMD const & v,
+    const fmat4x4SIMD &m
+)
+{
+    return v * inverse(m);
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD operator/
+(
+    const fmat4x4SIMD &m1,
+    const fmat4x4SIMD &m2
+)
+{
+    __m128 result[4];
+    __m128 inv[4];
+
+	sse_inverse_ps(&m2.Data[0].Data, inv);
+	sse_mul_ps(&m1.Data[0].Data, inv, result);
+
+    return fmat4x4SIMD(result);
+}
+
+
+//////////////////////////////////////////////////////////////
+// Unary constant operators
+GLM_FUNC_QUALIFIER fmat4x4SIMD const operator-
+(
+    fmat4x4SIMD const & m
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        -m[0],
+        -m[1],
+        -m[2],
+        -m[3]
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD const operator--
+(
+    fmat4x4SIMD const & m,
+    int
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] - 1.0f,
+        m[1] - 1.0f,
+        m[2] - 1.0f,
+        m[3] - 1.0f
+    );
+}
+
+GLM_FUNC_QUALIFIER fmat4x4SIMD const operator++
+(
+    fmat4x4SIMD const & m,
+    int
+)
+{
+    return detail::fmat4x4SIMD
+    (
+        m[0] + 1.0f,
+        m[1] + 1.0f,
+        m[2] + 1.0f,
+        m[3] + 1.0f
+    );
+}
+
 }//namespace detail
 
 GLM_FUNC_QUALIFIER detail::tmat4x4<float> mat4_cast

glm/gtx/simd_vec4.hpp

@@ -54,6 +54,14 @@
 #	pragma message("GLM: GLM_GTX_simd_vec4 extension included")
 #endif
 
+
+// Warning silencer for nameless struct/union.
+#if (GLM_COMPILER & GLM_COMPILER_VC)
+#   pragma warning(push)
+#   pragma warning(disable:4201)   // warning C4201: nonstandard extension used : nameless struct/union
+#endif
+
+
 namespace glm{
 namespace detail
 {
@@ -69,7 +77,15 @@ namespace detail
 		typedef fvec4SIMD type;
 		typedef tvec4<bool> bool_type;
 
-		__m128 Data;
+#ifdef GLM_SIMD_ENABLE_XYZW_UNION
+        union
+        {
+		    __m128 Data;
+            struct {float x, y, z, w;};
+        };
+#else
+        __m128 Data;
+#endif
 
 		//////////////////////////////////////
 		// Implicit basic constructors
@@ -490,6 +506,12 @@ namespace detail
 
 #include "simd_vec4.inl"
 
+
+#if (GLM_COMPILER & GLM_COMPILER_VC)
+#   pragma warning(pop)
+#endif
+
+
 #endif//(GLM_ARCH != GLM_ARCH_PURE)
 
 #endif//GLM_GTX_simd_vec4

glm/gtx/simd_vec4.inl

@@ -20,6 +20,9 @@ struct mask
 // Implicit basic constructors
 
 GLM_FUNC_QUALIFIER fvec4SIMD::fvec4SIMD()
+#ifdef GLM_SIMD_ENABLE_DEFAULT_INIT
+    : Data(_mm_set_ps(0.0f, 0.0f, 0.0f, 0.0f))
+#endif
 {}
 
 GLM_FUNC_QUALIFIER fvec4SIMD::fvec4SIMD(__m128 const & Data) :