Quaternion add and sub ops are not SSEd

HandmadeMath.h

@@ -1,5 +1,5 @@
 /*
-  HandmadeMath.h v1.7.0
+  HandmadeMath.h v1.6.0
   
   This is a single header file with a bunch of useful functions for game and
   graphics math operations.
@@ -172,10 +172,6 @@
           (*) Added array subscript operators for vector and matrix types in
               C++. This is provided as a convenience, but be aware that it may
               incur an extra function call in unoptimized builds.
-     1.7.0
-          (*) Renamed the 'Rows' member of hmm_mat4 to 'Columns'. Since our
-              matrices are column-major, this should have been named 'Columns'
-              from the start. 'Rows' is still present, but has been deprecated.
 
 
   LICENSE
@@ -451,10 +447,6 @@ typedef union hmm_mat4
     float Elements[4][4];
         
 #ifdef HANDMADE_MATH__USE_SSE
-    __m128 Columns[4];
-
-    // DEPRECATED. Our matrices are column-major, so this was named
-    // incorrectly. Use Columns instead.
     __m128 Rows[4];
 #endif
 
@@ -489,8 +481,17 @@ typedef union hmm_quaternion
         
         float W;
     };
+        
+    struct
+    {
+        hmm_vec4 XYZW;
+    };
     
     float Elements[4];
+        
+#if HANDMADE_MATH__USE_SSE
+    _m128 InternalElementsSSE;
+#endif        
 } hmm_quaternion;
 
 typedef int32_t hmm_bool;
@@ -1137,10 +1138,10 @@ HMM_INLINE hmm_vec4 HMM_NormalizeVec4(hmm_vec4 A)
 HMM_INLINE __m128 HMM_LinearCombineSSE(__m128 Left, hmm_mat4 Right)
 {
     __m128 Result;
-    Result = _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0x00), Right.Columns[0]);
-    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0x55), Right.Columns[1]));
-    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0xaa), Right.Columns[2]));
-    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0xff), Right.Columns[3]));
+    Result = _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0x00), Right.Rows[0]);
+    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0x55), Right.Rows[1]));
+    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0xaa), Right.Rows[2]));
+    Result = _mm_add_ps(Result, _mm_mul_ps(_mm_shuffle_ps(Left, Left, 0xff), Right.Rows[3]));
     
     return (Result);
 }
@@ -1175,7 +1176,7 @@ HMM_INLINE hmm_mat4 HMM_Transpose(hmm_mat4 Matrix)
 {
     hmm_mat4 Result = Matrix;
     
-    _MM_TRANSPOSE4_PS(Result.Columns[0], Result.Columns[1], Result.Columns[2], Result.Columns[3]);
+    _MM_TRANSPOSE4_PS(Result.Rows[0], Result.Rows[1], Result.Rows[2], Result.Rows[3]);
 
     return (Result);
 }
@@ -1188,10 +1189,10 @@ HMM_INLINE hmm_mat4 HMM_AddMat4(hmm_mat4 Left, hmm_mat4 Right)
 {
     hmm_mat4 Result;
 
-    Result.Columns[0] = _mm_add_ps(Left.Columns[0], Right.Columns[0]);
-    Result.Columns[1] = _mm_add_ps(Left.Columns[1], Right.Columns[1]);
-    Result.Columns[2] = _mm_add_ps(Left.Columns[2], Right.Columns[2]);
-    Result.Columns[3] = _mm_add_ps(Left.Columns[3], Right.Columns[3]);    
+    Result.Rows[0] = _mm_add_ps(Left.Rows[0], Right.Rows[0]);
+    Result.Rows[1] = _mm_add_ps(Left.Rows[1], Right.Rows[1]);
+    Result.Rows[2] = _mm_add_ps(Left.Rows[2], Right.Rows[2]);
+    Result.Rows[3] = _mm_add_ps(Left.Rows[3], Right.Rows[3]);    
 
     return (Result);
 }
@@ -1204,10 +1205,10 @@ HMM_INLINE hmm_mat4 HMM_SubtractMat4(hmm_mat4 Left, hmm_mat4 Right)
 {
     hmm_mat4 Result;
 
-    Result.Columns[0] = _mm_sub_ps(Left.Columns[0], Right.Columns[0]);
-    Result.Columns[1] = _mm_sub_ps(Left.Columns[1], Right.Columns[1]);
-    Result.Columns[2] = _mm_sub_ps(Left.Columns[2], Right.Columns[2]);
-    Result.Columns[3] = _mm_sub_ps(Left.Columns[3], Right.Columns[3]);
+    Result.Rows[0] = _mm_sub_ps(Left.Rows[0], Right.Rows[0]);
+    Result.Rows[1] = _mm_sub_ps(Left.Rows[1], Right.Rows[1]);
+    Result.Rows[2] = _mm_sub_ps(Left.Rows[2], Right.Rows[2]);
+    Result.Rows[3] = _mm_sub_ps(Left.Rows[3], Right.Rows[3]);
 
     return (Result);
 }
@@ -1223,10 +1224,10 @@ HMM_INLINE hmm_mat4 HMM_MultiplyMat4f(hmm_mat4 Matrix, float Scalar)
     hmm_mat4 Result;
 
     __m128 SSEScalar = _mm_set1_ps(Scalar);
-    Result.Columns[0] = _mm_mul_ps(Matrix.Columns[0], SSEScalar);
-    Result.Columns[1] = _mm_mul_ps(Matrix.Columns[1], SSEScalar);
-    Result.Columns[2] = _mm_mul_ps(Matrix.Columns[2], SSEScalar);
-    Result.Columns[3] = _mm_mul_ps(Matrix.Columns[3], SSEScalar);
+    Result.Rows[0] = _mm_mul_ps(Matrix.Rows[0], SSEScalar);
+    Result.Rows[1] = _mm_mul_ps(Matrix.Rows[1], SSEScalar);
+    Result.Rows[2] = _mm_mul_ps(Matrix.Rows[2], SSEScalar);
+    Result.Rows[3] = _mm_mul_ps(Matrix.Rows[3], SSEScalar);
 
     return (Result);
 }
@@ -1242,10 +1243,10 @@ HMM_INLINE hmm_mat4 HMM_DivideMat4f(hmm_mat4 Matrix, float Scalar)
     hmm_mat4 Result;
     
     __m128 SSEScalar = _mm_set1_ps(Scalar);
-    Result.Columns[0] = _mm_div_ps(Matrix.Columns[0], SSEScalar);
-    Result.Columns[1] = _mm_div_ps(Matrix.Columns[1], SSEScalar);
-    Result.Columns[2] = _mm_div_ps(Matrix.Columns[2], SSEScalar);
-    Result.Columns[3] = _mm_div_ps(Matrix.Columns[3], SSEScalar);    
+    Result.Rows[0] = _mm_div_ps(Matrix.Rows[0], SSEScalar);
+    Result.Rows[1] = _mm_div_ps(Matrix.Rows[1], SSEScalar);
+    Result.Rows[2] = _mm_div_ps(Matrix.Rows[2], SSEScalar);
+    Result.Rows[3] = _mm_div_ps(Matrix.Rows[3], SSEScalar);    
 
     return (Result);
 }
@@ -1325,11 +1326,14 @@ HMM_INLINE hmm_quaternion HMM_Quaternion(float X, float Y, float Z, float W)
 {
     hmm_quaternion Result;
 
+#if HANDMADE_MATH__USE_SSE
+    Result.InternalElementsSSE = _mm_setr_ps(X, Y, Z, W);
+#else 
     Result.X = X;
     Result.Y = Y;
     Result.Z = Z;
     Result.W = W;
-
+#endif
     return (Result);
 }
 
@@ -1337,11 +1341,15 @@ HMM_INLINE hmm_quaternion HMM_QuaternionV4(hmm_vec4 Vector)
 {
     hmm_quaternion Result;
 
+#if HANDMADE_MATH__USE_SSE        
+    Result.InternalElementsSSE = _mm_setr_ps(Vector.X, Vector.Y, Vector.Z, Vector.W);    
+#else
     Result.X = Vector.X;
     Result.Y = Vector.Y;
     Result.Z = Vector.Z;
-    Result.W = Vector.W;
-
+    Result.W = Vector.W;    
+#endif        
+        
     return (Result);
 }
 
@@ -1349,11 +1357,14 @@ HMM_INLINE hmm_quaternion HMM_AddQuaternion(hmm_quaternion Left, hmm_quaternion
 {
     hmm_quaternion Result;
 
+#if HANDMADE_MATH__USE_SSE        
+    Result.InternalElementsSSE = _mm_add_ps(Left.InternalElementsSSE, Right.InternalElementsSSE);        
+#else         
     Result.X = Left.X + Right.X;
     Result.Y = Left.Y + Right.Y;
     Result.Z = Left.Z + Right.Z;
     Result.W = Left.W + Right.W;
-
+#endif
     return (Result);
 }
 
@@ -1361,11 +1372,15 @@ HMM_INLINE hmm_quaternion HMM_SubtractQuaternion(hmm_quaternion Left, hmm_quater
 {
     hmm_quaternion Result;
 
+#if HANDMADE_MATH__USE_SSE
+    Result.InternalElementsSSE = _mm_sub_ps(Left.InternalElementsSSE, Right.InternalElementsSSE);
+#else
     Result.X = Left.X - Right.X;
     Result.Y = Left.Y - Right.Y;
     Result.Z = Left.Z - Right.Z;
     Result.W = Left.W - Right.W;
-
+#endif
+        
     return (Result);
 }
 
@@ -2260,10 +2275,10 @@ hmm_mat4 HMM_MultiplyMat4(hmm_mat4 Left, hmm_mat4 Right)
 
 #ifdef HANDMADE_MATH__USE_SSE
 
-    Result.Columns[0] = HMM_LinearCombineSSE(Right.Columns[0], Left);
-    Result.Columns[1] = HMM_LinearCombineSSE(Right.Columns[1], Left);
-    Result.Columns[2] = HMM_LinearCombineSSE(Right.Columns[2], Left);
-    Result.Columns[3] = HMM_LinearCombineSSE(Right.Columns[3], Left);     
+    Result.Rows[0] = HMM_LinearCombineSSE(Right.Rows[0], Left);
+    Result.Rows[1] = HMM_LinearCombineSSE(Right.Rows[1], Left);
+    Result.Rows[2] = HMM_LinearCombineSSE(Right.Rows[2], Left);
+    Result.Rows[3] = HMM_LinearCombineSSE(Right.Rows[3], Left);     
     
 #else
     int Columns;

README.md

@@ -10,7 +10,6 @@ To get started, go download [the latest release](https://github.com/HandmadeMath
 
 Version         | Changes        |
 ----------------|----------------|
-**1.7.0** | Renamed the 'Rows' member of hmm_mat4 to 'Columns'. Since our matrices are column-major, this should have been named 'Columns' from the start. 'Rows' is still present, but has been deprecated.
 **1.6.0** | Added array subscript operators for vector and matrix types in C++. This is provided as a convenience, but be aware that it may incur an extra function call in unoptimized builds.
 **1.5.1** | Fixed a bug with uninitialized elements in HMM_LookAt.
 **1.5.0** | Changed internal structure for better performance and inlining. As a result, `HANDMADE_MATH_NO_INLINE` has been removed and no longer has any effect.

test/categories/SSE.h

@@ -8,10 +8,10 @@ TEST(SSE, LinearCombine)
     hmm_mat4 MatrixTwo = HMM_Mat4d(4.0f);
     hmm_mat4 Result;
     
-    Result.Columns[0] = HMM_LinearCombineSSE(MatrixOne.Columns[0], MatrixTwo);
-    Result.Columns[1] = HMM_LinearCombineSSE(MatrixOne.Columns[1], MatrixTwo);
-    Result.Columns[2] = HMM_LinearCombineSSE(MatrixOne.Columns[2], MatrixTwo);
-    Result.Columns[3] = HMM_LinearCombineSSE(MatrixOne.Columns[3], MatrixTwo);
+    Result.Rows[0] = HMM_LinearCombineSSE(MatrixOne.Rows[0], MatrixTwo);
+    Result.Rows[1] = HMM_LinearCombineSSE(MatrixOne.Rows[1], MatrixTwo);
+    Result.Rows[2] = HMM_LinearCombineSSE(MatrixOne.Rows[2], MatrixTwo);
+    Result.Rows[3] = HMM_LinearCombineSSE(MatrixOne.Rows[3], MatrixTwo);
     
     {
         EXPECT_FLOAT_EQ(Result.Elements[0][0], 8.0f);
@@ -23,12 +23,14 @@ TEST(SSE, LinearCombine)
         EXPECT_FLOAT_EQ(Result.Elements[1][1], 8.0f);                
         EXPECT_FLOAT_EQ(Result.Elements[1][2], 0.0f);
         EXPECT_FLOAT_EQ(Result.Elements[1][3], 0.0f);
+
                         
         EXPECT_FLOAT_EQ(Result.Elements[2][0], 0.0f);
         EXPECT_FLOAT_EQ(Result.Elements[2][1], 0.0f);                
         EXPECT_FLOAT_EQ(Result.Elements[2][2], 8.0f);
         EXPECT_FLOAT_EQ(Result.Elements[2][3], 0.0f);
 
+        
         EXPECT_FLOAT_EQ(Result.Elements[3][0], 0.0f);
         EXPECT_FLOAT_EQ(Result.Elements[3][1], 0.0f);                
         EXPECT_FLOAT_EQ(Result.Elements[3][2], 0.0f);