Merge branch 'master' into simd

* this will help to implement zoom easily

README.md

@@ -25,6 +25,7 @@ you have the latest version
 - **[api rename]** by starting v0.4.5, **glm_simd** functions are renamed to **glmm_**  
 - **[new option]** by starting v0.4.5, you can disable alignment requirement, check options in docs.  
 - **[major change]** by starting v0.5.0, vec3 functions use **glm_vec3_** namespace, it was **glm_vec_** until v0.5.0
+- **[major change]** by starting v0.5.1, built-in alignment is removed from **vec3** and **mat3** types
 
 #### Note for C++ developers:
 If you don't aware about original GLM library yet, you may also want to look at:

configure.ac

@@ -7,7 +7,7 @@
 #*****************************************************************************
 
 AC_PREREQ([2.69])
-AC_INIT([cglm], [0.5.0], [info@recp.me])
+AC_INIT([cglm], [0.5.2], [info@recp.me])
 AM_INIT_AUTOMAKE([-Wall -Werror foreign subdir-objects])
 
 AC_CONFIG_MACRO_DIR([m4])

docs/source/cam.rst

@@ -36,6 +36,7 @@ Functions:
 #. :c:func:`glm_ortho_default`
 #. :c:func:`glm_ortho_default_s`
 #. :c:func:`glm_perspective`
+#. :c:func:`glm_persp_move_far`
 #. :c:func:`glm_perspective_default`
 #. :c:func:`glm_perspective_resize`
 #. :c:func:`glm_lookat`
@@ -145,6 +146,16 @@ Functions documentation
       | *[in]*  **farVal**  far clipping planes
       | *[out]* **dest**    result matrix
 
+.. c:function:: void  glm_persp_move_far(mat4 proj, float deltaFar)
+
+    | extend perspective projection matrix's far distance
+
+    | this function does not guarantee far >= near, be aware of that!
+
+    Parameters:
+      | *[in, out]*  **proj**      projection matrix to extend
+      | *[in]*       **deltaFar**  distance from existing far (negative to shink)
+
 .. c:function:: void glm_perspective_default(float aspect, mat4 dest)
 
      | set up perspective projection matrix with default near/far

docs/source/conf.py

@@ -62,9 +62,9 @@ author = u'Recep Aslantas'
 # built documents.
 #
 # The short X.Y version.
-version = u'0.5.0'
+version = u'0.5.2'
 # The full version, including alpha/beta/rc tags.
-release = u'0.5.0'
+release = u'0.5.2'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

docs/source/mat3.rst

@@ -29,6 +29,7 @@ Functions:
 #. :c:func:`glm_mat3_scale`
 #. :c:func:`glm_mat3_det`
 #. :c:func:`glm_mat3_inv`
+#. :c:func:`glm_mat3_trace`
 #. :c:func:`glm_mat3_swap_col`
 #. :c:func:`glm_mat3_swap_row`
 
@@ -133,6 +134,16 @@ Functions documentation
       | *[in]*  **mat**  matrix
       | *[out]* **dest** destination (inverse matrix)
 
+.. c:function:: void glm_mat3_trace(mat3 m)
+
+    | sum of the elements on the main diagonal from upper left to the lower right
+
+    Parameters:
+      | *[in]*  **m**  matrix
+
+    Returns:
+        trace of matrix
+
 .. c:function:: void  glm_mat3_swap_col(mat3 mat, int col1, int col2)
 
     swap two matrix columns

docs/source/mat4.rst

@@ -33,6 +33,8 @@ Functions:
 #. :c:func:`glm_mat4_mulN`
 #. :c:func:`glm_mat4_mulv`
 #. :c:func:`glm_mat4_mulv3`
+#. :c:func:`glm_mat3_trace`
+#. :c:func:`glm_mat3_trace3`
 #. :c:func:`glm_mat4_quat`
 #. :c:func:`glm_mat4_transpose_to`
 #. :c:func:`glm_mat4_transpose`
@@ -156,6 +158,27 @@ Functions documentation
     | *[in]*  **v**     vec3 (right, column vector)
     | *[out]* **dest**  vec3 (result, column vector)
 
+.. c:function:: void  glm_mat4_trace(mat4 m)
+
+    | sum of the elements on the main diagonal from upper left to the lower right
+
+    Parameters:
+      | *[in]*  **m**  matrix
+
+    Returns:
+        trace of matrix
+
+.. c:function:: void  glm_mat4_trace3(mat4 m)
+
+    | trace of matrix (rotation part)
+    | sum of the elements on the main diagonal from upper left to the lower right
+
+    Parameters:
+      | *[in]*  **m**  matrix
+
+    Returns:
+        trace of matrix
+
 .. c:function:: void  glm_mat4_quat(mat4 m, versor dest)
 
     convert mat4's rotation part to quaternion

docs/source/vec3.rst

@@ -39,7 +39,6 @@ Functions:
 #. :c:func:`glm_vec3_zero`
 #. :c:func:`glm_vec3_one`
 #. :c:func:`glm_vec3_dot`
-#. :c:func:`glm_vec3_cross`
 #. :c:func:`glm_vec3_norm2`
 #. :c:func:`glm_vec3_norm`
 #. :c:func:`glm_vec3_add`
@@ -65,6 +64,8 @@ Functions:
 #. :c:func:`glm_vec3_negate_to`
 #. :c:func:`glm_vec3_normalize`
 #. :c:func:`glm_vec3_normalize_to`
+#. :c:func:`glm_vec3_cross`
+#. :c:func:`glm_vec3_crossn`
 #. :c:func:`glm_vec3_distance2`
 #. :c:func:`glm_vec3_distance`
 #. :c:func:`glm_vec3_angle`
@@ -125,12 +126,21 @@ Functions documentation
 
 .. c:function:: void  glm_vec3_cross(vec3 a, vec3 b, vec3 d)
 
-    cross product
+    cross product of two vector (RH)
 
     Parameters:
-      | *[in]*  **a**  source 1
+      | *[in]*  **a**     vector 1
-      | *[in]*  **b**  source 2
+      | *[in]*  **b**     vector 2
-      | *[out]* **d**  destination
+      | *[out]* **dest**  destination
+
+.. c:function:: void  glm_vec3_crossn(vec3 a, vec3 b, vec3 dest)
+
+    cross product of two vector (RH) and normalize the result
+
+    Parameters:
+      | *[in]*  **a**     vector 1
+      | *[in]*  **b**     vector 2
+      | *[out]* **dest**  destination
 
 .. c:function:: float  glm_vec3_norm2(vec3 v)
 

include/cglm/call/cam.h

@@ -61,6 +61,10 @@ glmc_perspective(float fovy,
                  float farVal,
                  mat4 dest);
 
+CGLM_EXPORT
+void
+glmc_persp_move_far(mat4 proj, float deltaFar);
+
 CGLM_EXPORT
 void
 glmc_perspective_default(float aspect, mat4 dest);

include/cglm/call/mat3.h

@@ -44,6 +44,10 @@ CGLM_EXPORT
 void
 glmc_mat3_mulv(mat3 m, vec3 v, vec3 dest);
 
+CGLM_EXPORT
+float
+glmc_mat3_trace(mat3 m);
+
 CGLM_EXPORT
 void
 glmc_mat3_quat(mat3 m, versor dest);

include/cglm/call/mat4.h

@@ -61,6 +61,14 @@ CGLM_EXPORT
 void
 glmc_mat4_mulv3(mat4 m, vec3 v, float last, vec3 dest);
 
+CGLM_EXPORT
+float
+glmc_mat4_trace(mat4 m);
+
+CGLM_EXPORT
+float
+glmc_mat4_trace3(mat4 m);
+
 CGLM_EXPORT
 void
 glmc_mat4_quat(mat4 m, versor dest);

include/cglm/call/vec3.h

@@ -42,7 +42,11 @@ glmc_vec3_dot(vec3 a, vec3 b);
 
 CGLM_EXPORT
 void
-glmc_vec3_cross(vec3 a, vec3 b, vec3 d);
+glmc_vec3_cross(vec3 a, vec3 b, vec3 dest);
+
+CGLM_EXPORT
+void
+glmc_vec3_crossn(vec3 a, vec3 b, vec3 dest);
 
 CGLM_EXPORT
 float

include/cglm/cam.h

@@ -271,6 +271,30 @@ glm_perspective(float fovy,
   dest[3][2] = 2.0f * nearVal * farVal * fn;
 }
 
+/*!
+ * @brief extend perspective projection matrix's far distance
+ *
+ * this function does not guarantee far >= near, be aware of that!
+ *
+ * @param[in, out] proj      projection matrix to extend
+ * @param[in]      deltaFar  distance from existing far (negative to shink)
+ */
+CGLM_INLINE
+void
+glm_persp_move_far(mat4 proj, float deltaFar) {
+  float fn, farVal, nearVal, p22, p32;
+
+  p22        = proj[2][2];
+  p32        = proj[3][2];
+
+  nearVal    = p32 / (p22 - 1.0f);
+  farVal     = p32 / (p22 + 1.0f) + deltaFar;
+  fn         = 1.0f / (nearVal - farVal);
+
+  proj[2][2] = (nearVal + farVal) * fn;
+  proj[3][2] = 2.0f * nearVal * farVal * fn;
+}
+
 /*!
  * @brief set up perspective projection matrix with default near/far
  *        and angle values
@@ -323,9 +347,7 @@ glm_lookat(vec3 eye,
   glm_vec3_sub(center, eye, f);
   glm_vec3_normalize(f);
 
-  glm_vec3_cross(f, up, s);
+  glm_vec3_crossn(f, up, s);
-  glm_vec3_normalize(s);
-
   glm_vec3_cross(s, f, u);
 
   dest[0][0] = s[0];

include/cglm/mat3.h

@@ -21,6 +21,7 @@
    CGLM_INLINE void  glm_mat3_transpose_to(mat3 m, mat3 dest);
    CGLM_INLINE void  glm_mat3_transpose(mat3 m);
    CGLM_INLINE void  glm_mat3_mulv(mat3 m, vec3 v, vec3 dest);
+   CGLM_INLINE float glm_mat3_trace(mat3 m);
    CGLM_INLINE void  glm_mat3_scale(mat3 m, float s);
    CGLM_INLINE float glm_mat3_det(mat3 mat);
    CGLM_INLINE void  glm_mat3_inv(mat3 mat, mat3 dest);
@@ -207,6 +208,18 @@ glm_mat3_mulv(mat3 m, vec3 v, vec3 dest) {
   dest[2] = m[0][2] * v[0] + m[1][2] * v[1] + m[2][2] * v[2];
 }
 
+/*!
+ * @brief trace of matrix
+ *
+ * sum of the elements on the main diagonal from upper left to the lower right
+ *
+ * @param[in]  m matrix
+ */
+CGLM_INLINE
+float
+glm_mat3_trace(mat3 m) {
+  return m[0][0] + m[1][1] + m[2][2];
+}
 
 /*!
  * @brief convert mat3 to quaternion

include/cglm/mat4.h

@@ -29,6 +29,8 @@
    CGLM_INLINE void  glm_mat4_mulN(mat4 *matrices[], int len, mat4 dest);
    CGLM_INLINE void  glm_mat4_mulv(mat4 m, vec4 v, vec4 dest);
    CGLM_INLINE void  glm_mat4_mulv3(mat4 m, vec3 v, vec3 dest);
+   CGLM_INLINE float glm_mat4_trace(mat4 m);
+   CGLM_INLINE float glm_mat4_trace3(mat4 m);
    CGLM_INLINE void  glm_mat4_transpose_to(mat4 m, mat4 dest);
    CGLM_INLINE void  glm_mat4_transpose(mat4 m);
    CGLM_INLINE void  glm_mat4_scale_p(mat4 m, float s);
@@ -338,6 +340,32 @@ glm_mat4_mulv(mat4 m, vec4 v, vec4 dest) {
 #endif
 }
 
+/*!
+ * @brief trace of matrix
+ *
+ * sum of the elements on the main diagonal from upper left to the lower right
+ *
+ * @param[in]  m matrix
+ */
+CGLM_INLINE
+float
+glm_mat4_trace(mat4 m) {
+  return m[0][0] + m[1][1] + m[2][2] + m[3][3];
+}
+
+/*!
+ * @brief trace of matrix (rotation part)
+ *
+ * sum of the elements on the main diagonal from upper left to the lower right
+ *
+ * @param[in]  m matrix
+ */
+CGLM_INLINE
+float
+glm_mat4_trace3(mat4 m) {
+  return m[0][0] + m[1][1] + m[2][2];
+}
+
 /*!
  * @brief convert mat4's rotation part to quaternion
  *
@@ -476,7 +504,9 @@ glm_mat4_scale_p(mat4 m, float s) {
 CGLM_INLINE
 void
 glm_mat4_scale(mat4 m, float s) {
-#if defined( __SSE__ ) || defined( __SSE2__ )
+#ifdef __AVX__
+  glm_mat4_scale_avx(m, s);
+#elif defined( __SSE__ ) || defined( __SSE2__ )
   glm_mat4_scale_sse2(m, s);
 #else
   glm_mat4_scale_p(m, s);
@@ -526,7 +556,9 @@ glm_mat4_det(mat4 mat) {
 CGLM_INLINE
 void
 glm_mat4_inv(mat4 mat, mat4 dest) {
-#if defined( __SSE__ ) || defined( __SSE2__ )
+#ifdef __AVX__
+  glm_mat4_inv_avx(mat, dest);
+#elif defined( __SSE__ ) || defined( __SSE2__ )
   glm_mat4_inv_sse2(mat, dest);
 #else
   float t[6];
@@ -587,7 +619,9 @@ glm_mat4_inv(mat4 mat, mat4 dest) {
 CGLM_INLINE
 void
 glm_mat4_inv_fast(mat4 mat, mat4 dest) {
-#if defined( __SSE__ ) || defined( __SSE2__ )
+#ifdef __AVX__
+  glm_mat4_inv_fast_avx(mat, dest);
+#elif defined( __SSE__ ) || defined( __SSE2__ )
   glm_mat4_inv_fast_sse2(mat, dest);
 #else
   glm_mat4_inv(mat, dest);

include/cglm/simd/avx/mat4.h

@@ -14,31 +14,47 @@
 
 #include <immintrin.h>
 
+CGLM_INLINE
+void
+glm_mat4_scale_avx(mat4 m, float s) {
+  __m256 y0;
+  y0 = _mm256_set1_ps(s);
+
+  glmm_store256(m[0], _mm256_mul_ps(y0, glmm_load256(m[0])));
+  glmm_store256(m[2], _mm256_mul_ps(y0, glmm_load256(m[2])));
+}
+
 CGLM_INLINE
 void
 glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
   /* D = R * L (Column-Major) */
 
-  __m256 y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
+  __m256  y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
+  __m256i yi0, yi1, yi2, yi3;
 
-  y0 = glmm_load256(m2[0]); /* h g f e d c b a */
+  y0  = glmm_load256(m2[0]); /* h g f e d c b a */
-  y1 = glmm_load256(m2[2]); /* p o n m l k j i */
+  y1  = glmm_load256(m2[2]); /* p o n m l k j i */
 
-  y2 = glmm_load256(m1[0]); /* h g f e d c b a */
+  y2  = glmm_load256(m1[0]); /* h g f e d c b a */
-  y3 = glmm_load256(m1[2]); /* p o n m l k j i */
+  y3  = glmm_load256(m1[2]); /* p o n m l k j i */
 
   /* 0x03: 0b00000011 */
-  y4 = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */
+  y4  = _mm256_permute2f128_ps(y2, y2, 0x03); /* d c b a h g f e */
-  y5 = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */
+  y5  = _mm256_permute2f128_ps(y3, y3, 0x03); /* l k j i p o n m */
+
+  yi0 = _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0);
+  yi1 = _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2);
+  yi2 = _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1);
+  yi3 = _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3);
 
   /* f f f f a a a a */
   /* h h h h c c c c */
   /* e e e e b b b b */
   /* g g g g d d d d */
-  y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y6 = _mm256_permutevar_ps(y0, yi0);
-  y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y7 = _mm256_permutevar_ps(y0, yi1);
-  y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y8 = _mm256_permutevar_ps(y0, yi2);
-  y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+  y9 = _mm256_permutevar_ps(y0, yi3);
 
   glmm_store256(dest[0],
                 _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
@@ -50,10 +66,10 @@ glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
   /* p p p p k k k k */
   /* m m m m j j j j */
   /* o o o o l l l l */
-  y6 = _mm256_permutevar_ps(y1, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y6 = _mm256_permutevar_ps(y1, yi0);
-  y7 = _mm256_permutevar_ps(y1, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
+  y7 = _mm256_permutevar_ps(y1, yi1);
-  y8 = _mm256_permutevar_ps(y1, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y8 = _mm256_permutevar_ps(y1, yi2);
-  y9 = _mm256_permutevar_ps(y1, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
+  y9 = _mm256_permutevar_ps(y1, yi3);
 
   glmm_store256(dest[2],
                 _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
@@ -62,5 +78,365 @@ glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
                                             _mm256_mul_ps(y5, y9))));
 }
 
+CGLM_INLINE
+void
+glm_mat4_inv_avx(mat4 mat, mat4 dest) {
+  __m256  y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13;
+  __m256  yt0, yt1, yt2;
+  __m256  t0, t1, t2;
+  __m256  r1, r2;
+  __m256  flpsign;
+  __m256i yi1, yi2, yi3;
+
+  y0  = glmm_load256(mat[0]); /* h g f e d c b a */
+  y1  = glmm_load256(mat[2]); /* p o n m l k j i */
+
+  y2  = _mm256_permute2f128_ps(y1, y1, 0x00); /* l k j i l k j i */
+  y3  = _mm256_permute2f128_ps(y1, y1, 0x11); /* p o n m p o n m */
+  y4  = _mm256_permute2f128_ps(y0, y0, 0x03); /* d c b a h g f e */
+  y13 = _mm256_permute2f128_ps(y4, y4, 0x00); /* h g f e h g f e */
+
+  yi1     = _mm256_set_epi32(0, 0, 0, 0, 0, 1, 1, 2);
+  yi2     = _mm256_set_epi32(1, 1, 1, 2, 3, 2, 3, 3);
+  flpsign = _mm256_set_ps(0.f, -0.f, 0.f, -0.f, -0.f, 0.f, -0.f, 0.f);
+
+  /* i i i i i j j k */
+  /* n n n o p o p p */
+  /* m m m m m n n o */
+  /* j j j k l k l l */
+  /* e e e e e f f g */
+  /* f f f g h g h h */
+  y5  = _mm256_permutevar_ps(y2, yi1);
+  y6  = _mm256_permutevar_ps(y3, yi2);
+  y7  = _mm256_permutevar_ps(y3, yi1);
+  y8  = _mm256_permutevar_ps(y2, yi2);
+  y2  = _mm256_permutevar_ps(y13, yi1);
+  y3  = _mm256_permutevar_ps(y13, yi2);
+
+  yi1 = _mm256_set_epi32(2, 1, 0, 0, 2, 1, 0, 0);
+  yi2 = _mm256_set_epi32(2, 1, 1, 0, 2, 1, 1, 0);
+  yi3 = _mm256_set_epi32(3, 3, 2, 0, 3, 3, 2, 0);
+
+  /*
+   t0[0] = k * p - o * l;    t1[0] = g * p - o * h;    t2[0] = g * l - k * h;
+   t0[1] = j * p - n * l;    t1[1] = f * p - n * h;    t2[1] = f * l - j * h;
+   t0[2] = j * o - n * k;    t1[2] = f * o - n * g;    t2[2] = f * k - j * g;
+   t0[3] = i * p - m * l;    t1[3] = e * p - m * h;    t2[3] = e * l - i * h;
+   t0[4] = i * o - m * k;    t1[4] = e * o - m * g;    t2[4] = e * k - i * g;
+   t0[5] = i * n - m * j;    t1[5] = e * n - m * f;    t2[5] = e * j - i * f;
+   */
+  yt0 = _mm256_sub_ps(_mm256_mul_ps(y5, y6), _mm256_mul_ps(y7, y8));
+  yt1 = _mm256_sub_ps(_mm256_mul_ps(y2, y6), _mm256_mul_ps(y7, y3));
+  yt2 = _mm256_sub_ps(_mm256_mul_ps(y2, y8), _mm256_mul_ps(y5, y3));
+
+  /* t3 t2 t1 t0 t3 t2 t1 t0 */
+  /* t5 t5 t5 t4 t5 t5 t5 t4 */
+  y9  = _mm256_permute2f128_ps(yt0, yt0, 0x00);
+  y10 = _mm256_permute2f128_ps(yt0, yt0, 0x11);
+//
+  /* t2 t1 t0 t0 t2 t1 t0 t0 */
+  t0  = _mm256_permutevar_ps(y9, yi1);
+
+  /* t4 t3 t3 t1 t4 t3 t3 t1 */
+  y11 = _mm256_shuffle_ps(y9, y10, 0x4D);
+  y12 = _mm256_permutevar_ps(y11, yi2);
+  t1  = _mm256_permute2f128_ps(y12, y9, 0x00);
+
+  /* t5 t5 t4 t2 t5 t5 t4 t2 */
+  y11 = _mm256_shuffle_ps(y9, y10, 0x4A);
+  y12 = _mm256_permutevar_ps(y11, yi3);
+  t2  = _mm256_permute2f128_ps(y12, y12, 0x00);
+
+  /* a a a b e e e f */
+  /* b b c c f f g g */
+  /* c d d d g h h h */
+  y9  = _mm256_permute_ps(y4, 0x01);
+  y10 = _mm256_permute_ps(y4, 0x5A);
+  y11 = _mm256_permute_ps(y4, 0xBF);
+
+  /*
+   dest[0][0] =  f * t[0] - g * t[1] + h * t[2];
+   dest[1][0] =-(e * t[0] - g * t[3] + h * t[4]);
+   dest[2][0] =  e * t[1] - f * t[3] + h * t[5];
+   dest[3][0] =-(e * t[2] - f * t[4] + g * t[5]);
+
+   dest[0][1] =-(b * t[0] - c * t[1] + d * t[2]);
+   dest[1][1] =  a * t[0] - c * t[3] + d * t[4];
+   dest[2][1] =-(a * t[1] - b * t[3] + d * t[5]);
+   dest[3][1] =  a * t[2] - b * t[4] + c * t[5];
+   */
+  r1 = _mm256_xor_ps(_mm256_add_ps(_mm256_sub_ps(_mm256_mul_ps(y9, t0),
+                                                 _mm256_mul_ps(y10, t1)),
+                                   _mm256_mul_ps(y11, t2)),
+                     flpsign);
+
+  /* d c b a d c b a */
+  y2 = _mm256_permute2f128_ps(y0, y0, 0x0);
+
+  /* a a a b a a a b */
+  /* b b c c b b c c */
+  /* c d d d c d d d */
+  y3 = _mm256_permutevar_ps(y2, _mm256_set_epi32(0, 0, 0, 1, 0, 0, 0, 1));
+  y4 = _mm256_permutevar_ps(y2, _mm256_set_epi32(1, 1, 2, 2, 1, 1, 2, 2));
+  y5 = _mm256_permutevar_ps(y2, _mm256_set_epi32(2, 3, 3, 3, 2, 3, 3, 3));
+
+  /* t2[3] t2[2] t2[1] t2[0] t1[3] t1[2] t1[1] t1[0] */
+  /* t2[5] t2[5] t2[5] t2[4] t1[5] t1[5] t1[5] t1[4] */
+  y6 = _mm256_permute2f128_ps(yt1, yt2, 0x20);
+  y7 = _mm256_permute2f128_ps(yt1, yt2, 0x31);
+
+  /* t2[2] t2[1] t2[0] t2[0] t1[2] t1[1] t1[0] t1[0] */
+  t0 = _mm256_permutevar_ps(y6, yi1);
+
+  /* t1[4] t1[3] t1[3] t1[1] t1[4] t1[3] t1[3] t1[1] */
+
+  /* t1[4] t1[3] t1[3] t1[1] t1[4] t1[3] t1[3] t1[1] */
+  y11 = _mm256_shuffle_ps(y6, y7, 0x4D);
+  t1  = _mm256_permutevar_ps(y11, yi2);
+
+
+  /* t2[5] t2[5] t2[4] t2[2] t1[5] t1[5] t1[4] t1[2] */
+  y11 = _mm256_shuffle_ps(y6, y7, 0x4A);
+  t2  = _mm256_permutevar_ps(y11, yi3);
+
+  /*
+   dest[0][2] =  b * t1[0] - c * t1[1] + d * t1[2];
+   dest[1][2] =-(a * t1[0] - c * t1[3] + d * t1[4]);
+   dest[2][2] =  a * t1[1] - b * t1[3] + d * t1[5];
+   dest[3][2] =-(a * t1[2] - b * t1[4] + c * t1[5]);
+
+   dest[0][3] =-(b * t2[0] - c * t2[1] + d * t2[2]);
+   dest[1][3] =  a * t2[0] - c * t2[3] + d * t2[4];
+   dest[2][3] =-(a * t2[1] - b * t2[3] + d * t2[5]);
+   dest[3][3] =  a * t2[2] - b * t2[4] + c * t2[5];
+   */
+  r2 = _mm256_xor_ps(_mm256_add_ps(_mm256_sub_ps(_mm256_mul_ps(y3, t0),
+                                                 _mm256_mul_ps(y4, t1)),
+                                   _mm256_mul_ps(y5, t2)),
+                     flpsign);
+
+  /* determinant */
+
+  y4 = _mm256_mul_ps(y0, r1);
+  y4 = _mm256_permute2f128_ps(y4, y4, 0x30);
+  y4 = _mm256_dp_ps(y0, r1, 0xff);
+
+  y5 = _mm256_div_ps(_mm256_set1_ps(1.0f), y4);
+  r1 = _mm256_mul_ps(r1, y5);
+  r2 = _mm256_mul_ps(r2, y5);
+
+  /* transpose */
+
+  /* d c b a h g f e */
+  /* l k j i p o n m */
+  y0 = _mm256_permute2f128_ps(r1, r1, 0x03);
+  y1 = _mm256_permute2f128_ps(r2, r2, 0x03);
+
+  /* b a f e f e b a */
+  /* j i n m n m j i */
+  /* i m a e m i e a */
+  /* j n b f n j f b */
+  /* n j f b m i e a */
+  y2 = _mm256_shuffle_ps(r1, y0, 0x44);
+  y3 = _mm256_shuffle_ps(r2, y1, 0x44);
+  y4 = _mm256_shuffle_ps(y2, y3, 0x88);
+  y5 = _mm256_shuffle_ps(y2, y3, 0xDD);
+  y6 = _mm256_permute2f128_ps(y4, y5, 0x20);
+
+  /* d c h g h g d c */
+  /* l k p o p o l k */
+  /* k o c g o k g c */
+  /* l p d h p l h d */
+  /* p l h d o k g c */
+  y2 = _mm256_shuffle_ps(r1, y0, 0xEE);
+  y3 = _mm256_shuffle_ps(r2, y1, 0xEE);
+  y4 = _mm256_shuffle_ps(y2, y3, 0x88);
+  y5 = _mm256_shuffle_ps(y2, y3, 0xDD);
+  y7 = _mm256_permute2f128_ps(y4, y5, 0x20);
+
+  glmm_store256(dest[0], y6);
+  glmm_store256(dest[2], y7);
+}
+
+CGLM_INLINE
+void
+glm_mat4_inv_fast_avx(mat4 mat, mat4 dest) {
+  __m256  y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13;
+  __m256  yt0, yt1, yt2;
+  __m256  t0, t1, t2;
+  __m256  r1, r2;
+  __m256  flpsign;
+  __m256i yi1, yi2, yi3;
+
+  y0  = glmm_load256(mat[0]); /* h g f e d c b a */
+  y1  = glmm_load256(mat[2]); /* p o n m l k j i */
+
+  y2  = _mm256_permute2f128_ps(y1, y1, 0x00); /* l k j i l k j i */
+  y3  = _mm256_permute2f128_ps(y1, y1, 0x11); /* p o n m p o n m */
+  y4  = _mm256_permute2f128_ps(y0, y0, 0x03); /* d c b a h g f e */
+  y13 = _mm256_permute2f128_ps(y4, y4, 0x00); /* h g f e h g f e */
+
+  yi1     = _mm256_set_epi32(0, 0, 0, 0, 0, 1, 1, 2);
+  yi2     = _mm256_set_epi32(1, 1, 1, 2, 3, 2, 3, 3);
+  flpsign = _mm256_set_ps(0.f, -0.f, 0.f, -0.f, -0.f, 0.f, -0.f, 0.f);
+
+  /* i i i i i j j k */
+  /* n n n o p o p p */
+  /* m m m m m n n o */
+  /* j j j k l k l l */
+  /* e e e e e f f g */
+  /* f f f g h g h h */
+  y5  = _mm256_permutevar_ps(y2, yi1);
+  y6  = _mm256_permutevar_ps(y3, yi2);
+  y7  = _mm256_permutevar_ps(y3, yi1);
+  y8  = _mm256_permutevar_ps(y2, yi2);
+  y2  = _mm256_permutevar_ps(y13, yi1);
+  y3  = _mm256_permutevar_ps(y13, yi2);
+
+  yi1 = _mm256_set_epi32(2, 1, 0, 0, 2, 1, 0, 0);
+  yi2 = _mm256_set_epi32(2, 1, 1, 0, 2, 1, 1, 0);
+  yi3 = _mm256_set_epi32(3, 3, 2, 0, 3, 3, 2, 0);
+
+  /*
+   t0[0] = k * p - o * l;    t1[0] = g * p - o * h;    t2[0] = g * l - k * h;
+   t0[1] = j * p - n * l;    t1[1] = f * p - n * h;    t2[1] = f * l - j * h;
+   t0[2] = j * o - n * k;    t1[2] = f * o - n * g;    t2[2] = f * k - j * g;
+   t0[3] = i * p - m * l;    t1[3] = e * p - m * h;    t2[3] = e * l - i * h;
+   t0[4] = i * o - m * k;    t1[4] = e * o - m * g;    t2[4] = e * k - i * g;
+   t0[5] = i * n - m * j;    t1[5] = e * n - m * f;    t2[5] = e * j - i * f;
+   */
+  yt0 = _mm256_sub_ps(_mm256_mul_ps(y5, y6), _mm256_mul_ps(y7, y8));
+  yt1 = _mm256_sub_ps(_mm256_mul_ps(y2, y6), _mm256_mul_ps(y7, y3));
+  yt2 = _mm256_sub_ps(_mm256_mul_ps(y2, y8), _mm256_mul_ps(y5, y3));
+
+  /* t3 t2 t1 t0 t3 t2 t1 t0 */
+  /* t5 t5 t5 t4 t5 t5 t5 t4 */
+  y9  = _mm256_permute2f128_ps(yt0, yt0, 0x00);
+  y10 = _mm256_permute2f128_ps(yt0, yt0, 0x11);
+
+  /* t2 t1 t0 t0 t2 t1 t0 t0 */
+  t0  = _mm256_permutevar_ps(y9, yi1);
+
+  /* t4 t3 t3 t1 t4 t3 t3 t1 */
+  y11 = _mm256_shuffle_ps(y9, y10, 0x4D);
+  y12 = _mm256_permutevar_ps(y11, yi2);
+  t1  = _mm256_permute2f128_ps(y12, y9, 0x00);
+
+  /* t5 t5 t4 t2 t5 t5 t4 t2 */
+  y11 = _mm256_shuffle_ps(y9, y10, 0x4A);
+  y12 = _mm256_permutevar_ps(y11, yi3);
+  t2  = _mm256_permute2f128_ps(y12, y12, 0x00);
+
+  /* a a a b e e e f */
+  /* b b c c f f g g */
+  /* c d d d g h h h */
+  y9  = _mm256_permute_ps(y4, 0x01);
+  y10 = _mm256_permute_ps(y4, 0x5A);
+  y11 = _mm256_permute_ps(y4, 0xBF);
+
+  /*
+   dest[0][0] =  f * t[0] - g * t[1] + h * t[2];
+   dest[1][0] =-(e * t[0] - g * t[3] + h * t[4]);
+   dest[2][0] =  e * t[1] - f * t[3] + h * t[5];
+   dest[3][0] =-(e * t[2] - f * t[4] + g * t[5]);
+
+   dest[0][1] =-(b * t[0] - c * t[1] + d * t[2]);
+   dest[1][1] =  a * t[0] - c * t[3] + d * t[4];
+   dest[2][1] =-(a * t[1] - b * t[3] + d * t[5]);
+   dest[3][1] =  a * t[2] - b * t[4] + c * t[5];
+   */
+  r1 = _mm256_xor_ps(_mm256_add_ps(_mm256_sub_ps(_mm256_mul_ps(y9, t0),
+                                                 _mm256_mul_ps(y10, t1)),
+                                   _mm256_mul_ps(y11, t2)),
+                     flpsign);
+
+  /* d c b a d c b a */
+  y2 = _mm256_permute2f128_ps(y0, y0, 0x0);
+
+  /* a a a b a a a b */
+  /* b b c c b b c c */
+  /* c d d d c d d d */
+  y3 = _mm256_permutevar_ps(y2, _mm256_set_epi32(0, 0, 0, 1, 0, 0, 0, 1));
+  y4 = _mm256_permutevar_ps(y2, _mm256_set_epi32(1, 1, 2, 2, 1, 1, 2, 2));
+  y5 = _mm256_permutevar_ps(y2, _mm256_set_epi32(2, 3, 3, 3, 2, 3, 3, 3));
+
+  /* t2[3] t2[2] t2[1] t2[0] t1[3] t1[2] t1[1] t1[0] */
+  /* t2[5] t2[5] t2[5] t2[4] t1[5] t1[5] t1[5] t1[4] */
+  y6 = _mm256_permute2f128_ps(yt1, yt2, 0x20);
+  y7 = _mm256_permute2f128_ps(yt1, yt2, 0x31);
+
+  /* t2[2] t2[1] t2[0] t2[0] t1[2] t1[1] t1[0] t1[0] */
+  t0 = _mm256_permutevar_ps(y6, yi1);
+
+  /* t1[4] t1[3] t1[3] t1[1] t1[4] t1[3] t1[3] t1[1] */
+
+  /* t1[4] t1[3] t1[3] t1[1] t1[4] t1[3] t1[3] t1[1] */
+  y11 = _mm256_shuffle_ps(y6, y7, 0x4D);
+  t1  = _mm256_permutevar_ps(y11, yi2);
+
+
+  /* t2[5] t2[5] t2[4] t2[2] t1[5] t1[5] t1[4] t1[2] */
+  y11 = _mm256_shuffle_ps(y6, y7, 0x4A);
+  t2  = _mm256_permutevar_ps(y11, yi3);
+
+  /*
+   dest[0][2] =  b * t1[0] - c * t1[1] + d * t1[2];
+   dest[1][2] =-(a * t1[0] - c * t1[3] + d * t1[4]);
+   dest[2][2] =  a * t1[1] - b * t1[3] + d * t1[5];
+   dest[3][2] =-(a * t1[2] - b * t1[4] + c * t1[5]);
+
+   dest[0][3] =-(b * t2[0] - c * t2[1] + d * t2[2]);
+   dest[1][3] =  a * t2[0] - c * t2[3] + d * t2[4];
+   dest[2][3] =-(a * t2[1] - b * t2[3] + d * t2[5]);
+   dest[3][3] =  a * t2[2] - b * t2[4] + c * t2[5];
+   */
+  r2 = _mm256_xor_ps(_mm256_add_ps(_mm256_sub_ps(_mm256_mul_ps(y3, t0),
+                                                 _mm256_mul_ps(y4, t1)),
+                                   _mm256_mul_ps(y5, t2)),
+                     flpsign);
+
+  /* determinant */
+
+  y4 = _mm256_mul_ps(y0, r1);
+  y4 = _mm256_permute2f128_ps(y4, y4, 0x30);
+  y4 = _mm256_dp_ps(y0, r1, 0xff);
+
+  y5 = _mm256_rcp_ps(y4);
+  r1 = _mm256_mul_ps(r1, y5);
+  r2 = _mm256_mul_ps(r2, y5);
+
+  /* transpose */
+
+  /* d c b a h g f e */
+  /* l k j i p o n m */
+  y0 = _mm256_permute2f128_ps(r1, r1, 0x03);
+  y1 = _mm256_permute2f128_ps(r2, r2, 0x03);
+
+  /* b a f e f e b a */
+  /* j i n m n m j i */
+  /* i m a e m i e a */
+  /* j n b f n j f b */
+  /* n j f b m i e a */
+  y2 = _mm256_shuffle_ps(r1, y0, 0x44);
+  y3 = _mm256_shuffle_ps(r2, y1, 0x44);
+  y4 = _mm256_shuffle_ps(y2, y3, 0x88);
+  y5 = _mm256_shuffle_ps(y2, y3, 0xDD);
+  y6 = _mm256_permute2f128_ps(y4, y5, 0x20);
+
+  /* d c h g h g d c */
+  /* l k p o p o l k */
+  /* k o c g o k g c */
+  /* l p d h p l h d */
+  /* p l h d o k g c */
+  y2 = _mm256_shuffle_ps(r1, y0, 0xEE);
+  y3 = _mm256_shuffle_ps(r2, y1, 0xEE);
+  y4 = _mm256_shuffle_ps(y2, y3, 0x88);
+  y5 = _mm256_shuffle_ps(y2, y3, 0xDD);
+  y7 = _mm256_permute2f128_ps(y4, y5, 0x20);
+
+  glmm_store256(dest[0], y6);
+  glmm_store256(dest[2], y7);
+}
+
 #endif
 #endif /* cglm_mat_simd_avx_h */

include/cglm/types.h

@@ -10,12 +10,12 @@
 
 #if defined(_MSC_VER)
 /* do not use alignment for older visual studio versions */
-#if _MSC_VER < 1913 /*  Visual Studio 2017 version 15.6  */
+#  if _MSC_VER < 1913 /*  Visual Studio 2017 version 15.6  */
-#  define CGLM_ALL_UNALIGNED
+#    define CGLM_ALL_UNALIGNED
-#  define CGLM_ALIGN(X) /* no alignment */
+#    define CGLM_ALIGN(X) /* no alignment */
-#else
+#  else
-#  define CGLM_ALIGN(X) __declspec(align(X))
+#    define CGLM_ALIGN(X) __declspec(align(X))
-#endif
+#  endif
 #else
 #  define CGLM_ALIGN(X) __attribute((aligned(X)))
 #endif
@@ -33,20 +33,18 @@
 #endif
 
 typedef float                   vec2[2];
-typedef CGLM_ALIGN_IF(8)  float vec3[3];
+typedef float                   vec3[3];
 typedef int                    ivec3[3];
 typedef CGLM_ALIGN_IF(16) float vec4[4];
+typedef vec4                    versor;
+typedef vec3                    mat3[3];
 
 #ifdef __AVX__
-typedef CGLM_ALIGN_IF(32) vec3  mat3[3];
 typedef CGLM_ALIGN_IF(32) vec4  mat4[4];
 #else
-typedef                   vec3  mat3[3];
 typedef CGLM_ALIGN_IF(16) vec4  mat4[4];
 #endif
 
-typedef vec4                    versor;
-
 #define GLM_E         2.71828182845904523536028747135266250   /* e           */
 #define GLM_LOG2E     1.44269504088896340735992468100189214   /* log2(e)     */
 #define GLM_LOG10E    0.434294481903251827651128918916605082  /* log10(e)    */

include/cglm/vec3.h

@@ -21,7 +21,6 @@
    CGLM_INLINE void  glm_vec3_zero(vec3 v);
    CGLM_INLINE void  glm_vec3_one(vec3 v);
    CGLM_INLINE float glm_vec3_dot(vec3 a, vec3 b);
-   CGLM_INLINE void  glm_vec3_cross(vec3 a, vec3 b, vec3 d);
    CGLM_INLINE float glm_vec3_norm2(vec3 v);
    CGLM_INLINE float glm_vec3_norm(vec3 v);
    CGLM_INLINE void  glm_vec3_add(vec3 a, vec3 b, vec3 dest);
@@ -47,6 +46,8 @@
    CGLM_INLINE void  glm_vec3_inv_to(vec3 v, vec3 dest);
    CGLM_INLINE void  glm_vec3_normalize(vec3 v);
    CGLM_INLINE void  glm_vec3_normalize_to(vec3 v, vec3 dest);
+   CGLM_INLINE void  glm_vec3_cross(vec3 a, vec3 b, vec3 d);
+   CGLM_INLINE void  glm_vec3_crossn(vec3 a, vec3 b, vec3 dest);
    CGLM_INLINE float glm_vec3_distance(vec3 a, vec3 b);
    CGLM_INLINE float glm_vec3_angle(vec3 a, vec3 b);
    CGLM_INLINE void  glm_vec3_rotate(vec3 v, float angle, vec3 axis);
@@ -166,22 +167,6 @@ glm_vec3_dot(vec3 a, vec3 b) {
   return a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
 }
 
-/*!
- * @brief vec3 cross product
- *
- * @param[in]  a source 1
- * @param[in]  b source 2
- * @param[out] d destination
- */
-CGLM_INLINE
-void
-glm_vec3_cross(vec3 a, vec3 b, vec3 d) {
-  /* (u2.v3 - u3.v2, u3.v1 - u1.v3, u1.v2 - u2.v1) */
-  d[0] = a[1] * b[2] - a[2] * b[1];
-  d[1] = a[2] * b[0] - a[0] * b[2];
-  d[2] = a[0] * b[1] - a[1] * b[0];
-}
-
 /*!
  * @brief norm * norm (magnitude) of vec
  *
@@ -444,7 +429,7 @@ glm_vec3_maxadd(vec3 a, vec3 b, vec3 dest) {
  * it applies += operator so dest must be initialized
  *
  * @param[in]  a    vector
- * @param[in]  s    scalar
+ * @param[in]  b    scalar
  * @param[out] dest dest += min(a, b)
  */
 CGLM_INLINE
@@ -521,6 +506,36 @@ glm_vec3_normalize_to(vec3 v, vec3 dest) {
   glm_vec3_scale(v, 1.0f / norm, dest);
 }
 
+/*!
+ * @brief cross product of two vector (RH)
+ *
+ * @param[in]  a    vector 1
+ * @param[in]  b    vector 2
+ * @param[out] dest destination
+ */
+CGLM_INLINE
+void
+glm_vec3_cross(vec3 a, vec3 b, vec3 dest) {
+  /* (u2.v3 - u3.v2, u3.v1 - u1.v3, u1.v2 - u2.v1) */
+  dest[0] = a[1] * b[2] - a[2] * b[1];
+  dest[1] = a[2] * b[0] - a[0] * b[2];
+  dest[2] = a[0] * b[1] - a[1] * b[0];
+}
+
+/*!
+ * @brief cross product of two vector (RH) and normalize the result
+ *
+ * @param[in]  a    vector 1
+ * @param[in]  b    vector 2
+ * @param[out] dest destination
+ */
+CGLM_INLINE
+void
+glm_vec3_crossn(vec3 a, vec3 b, vec3 dest) {
+  glm_vec3_cross(a, b, dest);
+  glm_vec3_normalize(dest);
+}
+
 /*!
  * @brief angle betwen two vector
  *

include/cglm/vec4.h

@@ -331,9 +331,9 @@ glm_vec4_subs(vec4 v, float s, vec4 dest) {
 /*!
  * @brief multiply two vector (component-wise multiplication)
  *
- * @param a vector1
+ * @param a    vector1
- * @param b vector2
+ * @param b    vector2
- * @param d dest = (a[0] * b[0], a[1] * b[1], a[2] * b[2], a[3] * b[3])
+ * @param dest dest = (a[0] * b[0], a[1] * b[1], a[2] * b[2], a[3] * b[3])
  */
 CGLM_INLINE
 void
@@ -426,7 +426,6 @@ glm_vec4_divs(vec4 v, float s, vec4 dest) {
 #endif
 }
 
-
 /*!
  * @brief add two vectors and add result to sum
  *
@@ -553,7 +552,7 @@ glm_vec4_maxadd(vec4 a, vec4 b, vec4 dest) {
  * it applies += operator so dest must be initialized
  *
  * @param[in]  a    vector
- * @param[in]  s    scalar
+ * @param[in]  b    scalar
  * @param[out] dest dest += min(a, b)
  */
 CGLM_INLINE

include/cglm/version.h

@@ -10,6 +10,6 @@
 
 #define CGLM_VERSION_MAJOR 0
 #define CGLM_VERSION_MINOR 5
-#define CGLM_VERSION_PATCH 0
+#define CGLM_VERSION_PATCH 2
 
 #endif /* cglm_version_h */

src/cam.c

@@ -88,6 +88,12 @@ glmc_perspective(float fovy,
                   dest);
 }
 
+CGLM_EXPORT
+void
+glmc_persp_move_far(mat4 proj, float deltaFar) {
+  glm_persp_move_far(proj, deltaFar);
+}
+
 CGLM_EXPORT
 void
 glmc_perspective_default(float aspect, mat4 dest) {

src/mat3.c

@@ -50,6 +50,12 @@ glmc_mat3_mulv(mat3 m, vec3 v, vec3 dest) {
   glm_mat3_mulv(m, v, dest);
 }
 
+CGLM_EXPORT
+float
+glmc_mat3_trace(mat3 m) {
+  return glm_mat3_trace(m);
+}
+
 CGLM_EXPORT
 void
 glmc_mat3_quat(mat3 m, versor dest) {

src/mat4.c

@@ -74,6 +74,18 @@ glmc_mat4_mulv3(mat4 m, vec3 v, float last, vec3 dest) {
   glm_mat4_mulv3(m, v, last, dest);
 }
 
+CGLM_EXPORT
+float
+glmc_mat4_trace(mat4 m) {
+  return glm_mat4_trace(m);
+}
+
+CGLM_EXPORT
+float
+glmc_mat4_trace3(mat4 m) {
+  return glm_mat4_trace3(m);
+}
+
 CGLM_EXPORT
 void
 glmc_mat4_quat(mat4 m, versor dest) {

src/vec3.c

@@ -40,8 +40,14 @@ glmc_vec3_dot(vec3 a, vec3 b) {
 
 CGLM_EXPORT
 void
-glmc_vec3_cross(vec3 a, vec3 b, vec3 d) {
+glmc_vec3_cross(vec3 a, vec3 b, vec3 dest) {
-  glm_vec3_cross(a, b, d);
+  glm_vec3_cross(a, b, dest);
+}
+
+CGLM_EXPORT
+void
+glmc_vec3_crossn(vec3 a, vec3 b, vec3 dest) {
+  glm_vec3_crossn(a, b, dest);
 }
 
 CGLM_EXPORT

test/src/test_mat3.c

@@ -24,9 +24,9 @@ test_mat3(void **state) {
   for (i = 0; i < m; i++) {
     for (j = 0; j < n; j++) {
       if (i == j)
-        assert_true(m3[i][j] == 1.0f);
+        assert_true(glm_eq(m3[i][j], 1.0f));
       else
-        assert_true(m3[i][j] == 0.0f);
+        assert_true(glm_eq(m3[i][j], 0.0f));
     }
   }
 

test/src/test_mat4.c

@@ -24,9 +24,9 @@ test_mat4(void **state) {
   for (i = 0; i < m; i++) {
     for (j = 0; j < n; j++) {
       if (i == j)
-        assert_true(m3[i][j] == 1.0f);
+        assert_true(glm_eq(m3[i][j], 1.0f));
       else
-        assert_true(m3[i][j] == 0.0f);
+        assert_true(glm_eq(m3[i][j], 0.0f));
     }
   }
 

test/src/test_quat.c

@@ -25,7 +25,7 @@ test_quat(void **state) {
 
   /* 0. test identiy quat */
   glm_quat_identity(q4);
-  assert_true(glm_quat_real(q4) == cosf(glm_rad(0.0f) * 0.5f));
+  assert_true(glm_eq(glm_quat_real(q4), cosf(glm_rad(0.0f) * 0.5f)));
   glm_quat_mat4(q4, rot1);
   test_assert_mat4_eq2(rot1, GLM_MAT4_IDENTITY, 0.000009);
 
@@ -118,7 +118,7 @@ test_quat(void **state) {
   /* 9. test imag, real */
 
   /* 9.1 real */
-  assert_true(glm_quat_real(q4) == cosf(glm_rad(-90.0f) * 0.5f));
+  assert_true(glm_eq(glm_quat_real(q4), cosf(glm_rad(-90.0f) * 0.5f)));
 
   /* 9.1 imag */
   glm_quat_imag(q4, imag);