Merge pull request #63 from recp/avx

avx: replace binary constants (resolve https://github.com/recp/cglm/issues/62) with hex and fix glm_mul_avx

configure.ac

@@ -7,7 +7,7 @@
 #*****************************************************************************
 
 AC_PREREQ([2.69])
-AC_INIT([cglm], [0.4.8], [info@recp.me])
+AC_INIT([cglm], [0.4.9], [info@recp.me])
 AM_INIT_AUTOMAKE([-Wall -Werror foreign subdir-objects])
 
 AC_CONFIG_MACRO_DIR([m4])

docs/source/conf.py

@@ -62,9 +62,9 @@ author = u'Recep Aslantas'
 # built documents.
 #
 # The short X.Y version.
-version = u'0.4.8'
+version = u'0.4.9'
 # The full version, including alpha/beta/rc tags.
-release = u'0.4.8'
+release = u'0.4.9'
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

docs/source/mat3.rst

@@ -20,6 +20,7 @@ Functions:
 
 1. :c:func:`glm_mat3_copy`
 #. :c:func:`glm_mat3_identity`
+#. :c:func:`glm_mat3_identity_array`
 #. :c:func:`glm_mat3_mul`
 #. :c:func:`glm_mat3_transpose_to`
 #. :c:func:`glm_mat3_transpose`
@@ -49,6 +50,14 @@ Functions documentation
     Parameters:
       | *[out]* **mat**  matrix
 
+.. c:function:: void  glm_mat3_identity_array(mat3 * __restrict mat, size_t count)
+
+    make given matrix array's each element identity matrix
+
+    Parameters:
+      | *[in,out]* **mat**  matrix array (must be aligned (16/32) if alignment is not disabled)
+      | *[in]* **count**  count of matrices
+
 .. c:function:: void  glm_mat3_mul(mat3 m1, mat3 m2, mat3 dest)
 
     multiply m1 and m2 to dest

docs/source/mat4.rst

@@ -25,6 +25,7 @@ Functions:
 1. :c:func:`glm_mat4_ucopy`
 #. :c:func:`glm_mat4_copy`
 #. :c:func:`glm_mat4_identity`
+#. :c:func:`glm_mat4_identity_array`
 #. :c:func:`glm_mat4_pick3`
 #. :c:func:`glm_mat4_pick3t`
 #. :c:func:`glm_mat4_ins3`
@@ -69,6 +70,14 @@ Functions documentation
     Parameters:
       | *[out]* **mat**  matrix
 
+.. c:function:: void  glm_mat4_identity_array(mat4 * __restrict mat, size_t count)
+
+    make given matrix array's each element identity matrix
+
+    Parameters:
+      | *[in,out]* **mat**  matrix array (must be aligned (16/32) if alignment is not disabled)
+      | *[in]* **count**  count of matrices
+
 .. c:function:: void  glm_mat4_pick3(mat4 mat, mat3 dest)
 
     copy upper-left of mat4 to mat3

docs/source/quat.rst

@@ -27,6 +27,7 @@ Macros:
 Functions:
 
 1. :c:func:`glm_quat_identity`
+#. :c:func:`glm_quat_identity_array`
 #. :c:func:`glm_quat_init`
 #. :c:func:`glm_quat`
 #. :c:func:`glm_quatv`
@@ -70,6 +71,14 @@ Functions documentation
     Parameters:
       | *[in, out]*  **q**    quaternion
 
+.. c:function:: void  glm_quat_identity_array(versor * __restrict q, size_t count)
+
+    | make given quaternion array's each element identity quaternion
+
+    Parameters:
+      | *[in, out]*  **q**   quat array (must be aligned (16) if alignment is not disabled)
+      | *[in]*  **count**    count of quaternions
+
 .. c:function:: void  glm_quat_init(versor q, float x, float y, float z, float w)
 
     | inits quaternion with given values

include/cglm/affine-mat.h

@@ -151,7 +151,7 @@ glm_inv_tr(mat4 mat) {
 #if defined( __SSE__ ) || defined( __SSE2__ )
   glm_inv_tr_sse2(mat);
 #else
-  CGLM_ALIGN(16) mat3 r;
+  CGLM_ALIGN_MAT mat3 r;
   CGLM_ALIGN(16) vec3 t;
 
   /* rotate */

include/cglm/affine.h

@@ -244,7 +244,7 @@ glm_scale_uni(mat4 m, float s) {
 CGLM_INLINE
 void
 glm_rotate_x(mat4 m, float angle, mat4 dest) {
-  CGLM_ALIGN(16) mat4 t = GLM_MAT4_IDENTITY_INIT;
+  CGLM_ALIGN_MAT mat4 t = GLM_MAT4_IDENTITY_INIT;
   float c, s;
 
   c = cosf(angle);
@@ -269,7 +269,7 @@ glm_rotate_x(mat4 m, float angle, mat4 dest) {
 CGLM_INLINE
 void
 glm_rotate_y(mat4 m, float angle, mat4 dest) {
-  CGLM_ALIGN(16) mat4 t = GLM_MAT4_IDENTITY_INIT;
+  CGLM_ALIGN_MAT mat4 t = GLM_MAT4_IDENTITY_INIT;
   float c, s;
 
   c = cosf(angle);
@@ -294,7 +294,7 @@ glm_rotate_y(mat4 m, float angle, mat4 dest) {
 CGLM_INLINE
 void
 glm_rotate_z(mat4 m, float angle, mat4 dest) {
-  CGLM_ALIGN(16) mat4 t = GLM_MAT4_IDENTITY_INIT;
+  CGLM_ALIGN_MAT mat4 t = GLM_MAT4_IDENTITY_INIT;
   float c, s;
 
   c = cosf(angle);
@@ -351,7 +351,7 @@ glm_rotate_make(mat4 m, float angle, vec3 axis) {
 CGLM_INLINE
 void
 glm_rotate(mat4 m, float angle, vec3 axis) {
-  CGLM_ALIGN(16) mat4 rot;
+  CGLM_ALIGN_MAT mat4 rot;
   glm_rotate_make(rot, angle, axis);
   glm_mul_rot(m, rot, m);
 }

include/cglm/bezier.h

@@ -1,124 +0,0 @@
-/*
- * Copyright (c), Recep Aslantas.
- *
- * MIT License (MIT), http://opensource.org/licenses/MIT
- * Full license can be found in the LICENSE file
- */
-
-#ifndef cglm_bezier_h
-#define cglm_bezier_h
-
-#include "common.h"
-
-/*!
- * @brief cubic bezier interpolation
- *
- * @param[in]  t    time between 0 and 1
- * @param[in]  p0   begin point
- * @param[in]  c0   control point 1
- * @param[in]  c1   control point 2
- * @param[in]  p1   end point
- *
- * @return B(s)
- */
-CGLM_INLINE
-float
-glm_bezier_cubic(float t, float p0, float c0, float c1, float p1) {
-  float s, ss, tt;
-
-  s  = 1.0f - t;
-  ss = s * s;
-  tt = t * t;
-
-  return p0 * ss * s + 3.0f * c0 * t * ss + 3.0f * c1 * tt * s + p1 * tt * t;
-}
-
-/*!
- * @brief cubic bezier interpolation for vec3
- *
- * @param[in]  t    time between 0 and 1
- * @param[in]  p0   begin point
- * @param[in]  c0   control point 1
- * @param[in]  c1   control point 2
- * @param[in]  p1   end point
- * @param[in]  dest destination
- */
-CGLM_INLINE
-void
-glm_bezier_cubicv3(float t, vec3 p0, vec3 c0, vec3 c1, vec3 p1, vec3 dest) {
-  vec3  Bs, tmp0, tmp1, tmp2;
-  float s, ss, tt;
-
-  s  = 1.0f - t;
-  ss = s * s;
-  tt = t * t;
-
-  glm_vec_scale(p0, ss * s, Bs);
-  glm_vec_scale(c0, 3.0f * t * ss, tmp0);
-  glm_vec_scale(c1, 3.0f * tt * s, tmp1);
-  glm_vec_scale(p1, tt * t, tmp2);
-
-  glm_vec_add(Bs, tmp0, Bs);
-  glm_vec_add(Bs, tmp1, Bs);
-  glm_vec_add(Bs, tmp2, Bs);
-
-  glm_vec_copy(Bs, dest);
-}
-
-/*!
- * @brief cubic hermite interpolation
- *
- * @param[in]  t    time between 0 and 1
- * @param[in]  p0   begin point
- * @param[in]  t0   tangent 1
- * @param[in]  t1   tangent 2
- * @param[in]  p1   end point
- *
- * @return B(s)
- */
-CGLM_INLINE
-float
-glm_hermite_cubic(float t, float p0, float t0, float t1, float p1) {
-  float tt, ttt;
-
-  tt  = t * t;
-  ttt = tt * t;
-
-  return p0 * (2.0f * ttt - 3.0f * tt + 1)
-          + t0 * (ttt - 2.0f * tt + t)
-          + p1 * (3.0f * tt - 2.0f * ttt)
-          - t1 * (ttt - tt);
-}
-
-/*!
- * @brief cubic hermite interpolation for vec3
- *
- * @param[in]  t    time between 0 and 1
- * @param[in]  p0   begin point
- * @param[in]  t0   tangent 1
- * @param[in]  t1   tangent 2
- * @param[in]  p1   end point
- * @param[in]  dest destination
- */
-CGLM_INLINE
-void
-glm_hermite_cubicv3(float t, vec3 p0, vec3 t0, vec3 t1, vec3 p1, vec3 dest) {
-  vec3  Hs, tmp0, tmp1, tmp2;
-  float tt, ttt;
-
-  tt  = t * t;
-  ttt = tt * t;
-
-  glm_vec_scale(p0, 2.0f * ttt - 3.0f * tt + 1, Hs);
-  glm_vec_scale(t0, ttt - 2.0f * tt + t, tmp0);
-  glm_vec_scale(p1, 3.0f * tt - 2.0f * ttt, tmp1);
-  glm_vec_scale(t1, ttt - tt, tmp2);
-
-  glm_vec_add(Hs, tmp0, Hs);
-  glm_vec_add(Hs, tmp1, Hs);
-  glm_vec_add(Hs, tmp2, Hs);
-
-  glm_vec_copy(Hs, dest);
-}
-
-#endif /* cglm_bezier_h */

include/cglm/call/mat3.h

@@ -24,6 +24,10 @@ CGLM_EXPORT
 void
 glmc_mat3_identity(mat3 mat);
 
+CGLM_EXPORT
+void
+glmc_mat3_identity_array(mat3 * __restrict mat, size_t count);
+
 CGLM_EXPORT
 void
 glmc_mat3_mul(mat3 m1, mat3 m2, mat3 dest);

include/cglm/call/mat4.h

@@ -29,6 +29,10 @@ CGLM_EXPORT
 void
 glmc_mat4_identity(mat4 mat);
 
+CGLM_EXPORT
+void
+glmc_mat4_identity_array(mat4 * __restrict mat, size_t count);
+
 CGLM_EXPORT
 void
 glmc_mat4_pick3(mat4 mat, mat3 dest);

include/cglm/call/quat.h

@@ -17,6 +17,10 @@ CGLM_EXPORT
 void
 glmc_quat_identity(versor q);
 
+CGLM_EXPORT
+void
+glmc_quat_identity_array(versor * __restrict q, size_t count);
+
 CGLM_EXPORT
 void
 glmc_quat_init(versor q, float x, float y, float z, float w);

include/cglm/cam.h

@@ -198,26 +198,15 @@ glm_ortho_aabb_pz(vec3 box[2], float padding, mat4 dest) {
  */
 CGLM_INLINE
 void
-glm_ortho_default(float aspect,
-                  mat4  dest) {
+glm_ortho_default(float aspect, mat4 dest) {
   if (aspect >= 1.0f) {
-    glm_ortho(-1.0f * aspect,
-               1.0f * aspect,
-              -1.0f,
-               1.0f,
-              -100.0f,
-               100.0f,
-               dest);
-	  return;
+    glm_ortho(-aspect, aspect, -1.0f, 1.0f, -100.0f, 100.0f, dest);
+    return;
   }
 
-  glm_ortho(-1.0f,
-             1.0f,
-            -1.0f / aspect,
-             1.0f / aspect,
-            -100.0f,
-             100.0f,
-             dest);
+  aspect = 1.0f / aspect;
+
+  glm_ortho(-1.0f, 1.0f, -aspect, aspect, -100.0f, 100.0f, dest);
 }
 
 /*!
@@ -240,7 +229,7 @@ glm_ortho_default_s(float aspect,
               -size - 100.0f,
                size + 100.0f,
                dest);
-	  return;
+    return;
   }
 
   glm_ortho(-size,
@@ -291,13 +280,8 @@ glm_perspective(float fovy,
  */
 CGLM_INLINE
 void
-glm_perspective_default(float aspect,
-                        mat4  dest) {
-  glm_perspective((float)CGLM_PI_4,
-                  aspect,
-                  0.01f,
-                  100.0f,
-                  dest);
+glm_perspective_default(float aspect, mat4 dest) {
+  glm_perspective(GLM_PI_4f, aspect, 0.01f, 100.0f, dest);
 }
 
 /*!
@@ -310,8 +294,7 @@ glm_perspective_default(float aspect,
  */
 CGLM_INLINE
 void
-glm_perspective_resize(float aspect,
-                       mat4  proj) {
+glm_perspective_resize(float aspect, mat4 proj) {
   if (proj[0][0] == 0.0f)
     return;
 

include/cglm/cglm.h

@@ -26,6 +26,5 @@
 #include "project.h"
 #include "sphere.h"
 #include "ease.h"
-#include "bezier.h"
 
 #endif /* cglm_h */

include/cglm/ease.h

@@ -19,19 +19,19 @@ glm_ease_linear(float t) {
 CGLM_INLINE
 float
 glm_ease_sine_in(float t) {
-  return sinf((t - 1.0f) * CGLM_PI_2) + 1.0f;
+  return sinf((t - 1.0f) * GLM_PI_2f) + 1.0f;
 }
 
 CGLM_INLINE
 float
 glm_ease_sine_out(float t) {
-  return sinf(t * CGLM_PI_2);
+  return sinf(t * GLM_PI_2f);
 }
 
 CGLM_INLINE
 float
 glm_ease_sine_inout(float t) {
-  return 0.5f * (1.0f - cosf(t * CGLM_PI));
+  return 0.5f * (1.0f - cosf(t * GLM_PIf));
 }
 
 CGLM_INLINE
@@ -254,13 +254,13 @@ glm_ease_back_inout(float t) {
 CGLM_INLINE
 float
 glm_ease_elast_in(float t) {
-  return sinf(13.0f * CGLM_PI_2 * t) * powf(2.0f, 10.0f * (t - 1.0f));
+  return sinf(13.0f * GLM_PI_2f * t) * powf(2.0f, 10.0f * (t - 1.0f));
 }
 
 CGLM_INLINE
 float
 glm_ease_elast_out(float t) {
-  return sinf(-13.0f * CGLM_PI_2 * (t + 1.0f)) * powf(2.0f, -10.0f * t) + 1.0f;
+  return sinf(-13.0f * GLM_PI_2f * (t + 1.0f)) * powf(2.0f, -10.0f * t) + 1.0f;
 }
 
 CGLM_INLINE
@@ -271,10 +271,10 @@ glm_ease_elast_inout(float t) {
   a = 2.0f * t;
 
   if (t < 0.5f)
-    return 0.5f * sinf(13.0f * CGLM_PI_2 * a)
+    return 0.5f * sinf(13.0f * GLM_PI_2f * a)
                 * powf(2.0f, 10.0f * (a - 1.0f));
 
-  return 0.5f * (sinf(-13.0f * CGLM_PI_2 * a)
+  return 0.5f * (sinf(-13.0f * GLM_PI_2f * a)
                  * powf(2.0f, -10.0f * (a - 1.0f)) + 2.0f);
 }
 

include/cglm/euler.h

@@ -84,12 +84,12 @@ glm_euler_angles(mat4 m, vec3 dest) {
       thetaZ = atan2f(-m10, m00);
     } else { /* m20 == -1 */
       /* Not a unique solution */
-      thetaY = -CGLM_PI_2;
+      thetaY = -GLM_PI_2f;
       thetaX = -atan2f(m01, m11);
       thetaZ =  0.0f;
     }
   } else { /* m20 == +1 */
-    thetaY = CGLM_PI_2;
+    thetaY = GLM_PI_2f;
     thetaX = atan2f(m01, m11);
     thetaZ = 0.0f;
   }

include/cglm/io.h

@@ -25,7 +25,7 @@
 
 CGLM_INLINE
 void
-glm_mat4_print(mat4   matrix,
+glm_mat4_print(mat4              matrix,
                FILE * __restrict ostream) {
   int i;
   int j;
@@ -55,7 +55,7 @@ glm_mat4_print(mat4   matrix,
 
 CGLM_INLINE
 void
-glm_mat3_print(mat3 matrix,
+glm_mat3_print(mat3              matrix,
                FILE * __restrict ostream) {
   int i;
   int j;
@@ -85,7 +85,7 @@ glm_mat3_print(mat3 matrix,
 
 CGLM_INLINE
 void
-glm_vec4_print(vec4 vec,
+glm_vec4_print(vec4              vec,
                FILE * __restrict ostream) {
   int i;
 
@@ -107,7 +107,7 @@ glm_vec4_print(vec4 vec,
 
 CGLM_INLINE
 void
-glm_vec3_print(vec3 vec,
+glm_vec3_print(vec3              vec,
                FILE * __restrict ostream) {
   int i;
 
@@ -129,7 +129,7 @@ glm_vec3_print(vec3 vec,
 
 CGLM_INLINE
 void
-glm_ivec3_print(ivec3 vec,
+glm_ivec3_print(ivec3             vec,
                 FILE * __restrict ostream) {
   int i;
 
@@ -151,7 +151,7 @@ glm_ivec3_print(ivec3 vec,
 
 CGLM_INLINE
 void
-glm_versor_print(versor vec,
+glm_versor_print(versor            vec,
                  FILE * __restrict ostream) {
   int i;
 

include/cglm/mat3.h

@@ -16,6 +16,7 @@
  Functions:
    CGLM_INLINE void  glm_mat3_copy(mat3 mat, mat3 dest);
    CGLM_INLINE void  glm_mat3_identity(mat3 mat);
+   CGLM_INLINE void  glm_mat3_identity_array(mat3 * restrict mat, size_t count);
    CGLM_INLINE void  glm_mat3_mul(mat3 m1, mat3 m2, mat3 dest);
    CGLM_INLINE void  glm_mat3_transpose_to(mat3 m, mat3 dest);
    CGLM_INLINE void  glm_mat3_transpose(mat3 m);
@@ -81,10 +82,29 @@ glm_mat3_copy(mat3 mat, mat3 dest) {
 CGLM_INLINE
 void
 glm_mat3_identity(mat3 mat) {
-  CGLM_ALIGN(16) mat3 t = GLM_MAT3_IDENTITY_INIT;
+  CGLM_ALIGN_MAT mat3 t = GLM_MAT3_IDENTITY_INIT;
   glm_mat3_copy(t, mat);
 }
 
+/*!
+ * @brief make given matrix array's each element identity matrix
+ *
+ * @param[in, out]  mat   matrix array (must be aligned (16/32)
+ *                        if alignment is not disabled)
+ *
+ * @param[in]       count count of matrices
+ */
+CGLM_INLINE
+void
+glm_mat3_identity_array(mat3 * __restrict mat, size_t count) {
+  CGLM_ALIGN_MAT mat3 t = GLM_MAT3_IDENTITY_INIT;
+  size_t i;
+
+  for (i = 0; i < count; i++) {
+    glm_mat3_copy(t, mat[i]);
+  }
+}
+
 /*!
  * @brief multiply m1 and m2 to dest
  *
@@ -155,7 +175,7 @@ glm_mat3_transpose_to(mat3 m, mat3 dest) {
 CGLM_INLINE
 void
 glm_mat3_transpose(mat3 m) {
-  CGLM_ALIGN(16) mat3 tmp;
+  CGLM_ALIGN_MAT mat3 tmp;
 
   tmp[0][1] = m[1][0];
   tmp[0][2] = m[2][0];

include/cglm/mat4.h

@@ -23,6 +23,7 @@
    CGLM_INLINE void  glm_mat4_ucopy(mat4 mat, mat4 dest);
    CGLM_INLINE void  glm_mat4_copy(mat4 mat, mat4 dest);
    CGLM_INLINE void  glm_mat4_identity(mat4 mat);
+   CGLM_INLINE void  glm_mat4_identity_array(mat4 * restrict mat, size_t count);
    CGLM_INLINE void  glm_mat4_pick3(mat4 mat, mat3 dest);
    CGLM_INLINE void  glm_mat4_pick3t(mat4 mat, mat3 dest);
    CGLM_INLINE void  glm_mat4_ins3(mat3 mat, mat4 dest);
@@ -139,10 +140,29 @@ glm_mat4_copy(mat4 mat, mat4 dest) {
 CGLM_INLINE
 void
 glm_mat4_identity(mat4 mat) {
-  CGLM_ALIGN(16) mat4 t = GLM_MAT4_IDENTITY_INIT;
+  CGLM_ALIGN_MAT mat4 t = GLM_MAT4_IDENTITY_INIT;
   glm_mat4_copy(t, mat);
 }
 
+/*!
+ * @brief make given matrix array's each element identity matrix
+ *
+ * @param[in, out]  mat   matrix array (must be aligned (16/32)
+ *                        if alignment is not disabled)
+ *
+ * @param[in]       count count of matrices
+ */
+CGLM_INLINE
+void
+glm_mat4_identity_array(mat4 * __restrict mat, size_t count) {
+  CGLM_ALIGN_MAT mat4 t = GLM_MAT4_IDENTITY_INIT;
+  size_t i;
+
+  for (i = 0; i < count; i++) {
+    glm_mat4_copy(t, mat[i]);
+  }
+}
+
 /*!
  * @brief copy upper-left of mat4 to mat3
  *

include/cglm/quat.h

@@ -103,6 +103,25 @@ glm_quat_identity(versor q) {
   glm_vec4_copy(v, q);
 }
 
+/*!
+ * @brief make given quaternion array's each element identity quaternion
+ *
+ * @param[in, out]  q     quat array (must be aligned (16)
+ *                        if alignment is not disabled)
+ *
+ * @param[in]       count count of quaternions
+ */
+CGLM_INLINE
+void
+glm_quat_identity_array(versor * __restrict q, size_t count) {
+  CGLM_ALIGN(16) versor v = GLM_QUAT_IDENTITY_INIT;
+  size_t i;
+
+  for (i = 0; i < count; i++) {
+    glm_vec4_copy(v, q[i]);
+  }
+}
+
 /*!
  * @brief inits quaterion with raw values
  *
@@ -670,7 +689,7 @@ glm_quat_for(vec3 dir, vec3 fwd, vec3 up, versor dest) {
 
   dot = glm_vec_dot(dir, fwd);
   if (fabsf(dot + 1.0f)  < 0.000001f) {
-    glm_quat_init(dest, up[0], up[1], up[2], CGLM_PI);
+    glm_quat_init(dest, up[0], up[1], up[2], GLM_PIf);
     return;
   }
 
@@ -742,7 +761,7 @@ glm_quat_rotatev(versor q, vec3 v, vec3 dest) {
 CGLM_INLINE
 void
 glm_quat_rotate(mat4 m, versor q, mat4 dest) {
-  CGLM_ALIGN(16) mat4 rot;
+  CGLM_ALIGN_MAT mat4 rot;
   glm_quat_mat4(q, rot);
   glm_mul_rot(m, rot, dest);
 }

include/cglm/simd/avx/affine.h

@@ -27,21 +27,24 @@ glm_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
   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 */
 
-  y4 = _mm256_permute2f128_ps(y2, y2, 0b00000011); /* d c b a h g f e */
-  y5 = _mm256_permute2f128_ps(y3, y3, 0b00000000); /* l k j i l k j i */
+  /* 0x03: 0b00000011 */
+  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 */
 
   /* f f f f a a a a */
-  /* g g g g c c c c */
+  /* h h h h c c c c */
   /* e e e e b b b b */
-  y7 = _mm256_permute_ps(y0, 0b10101010);
+  /* g g g g d d d d */
   y6 = _mm256_permutevar_ps(y0, _mm256_set_epi32(1, 1, 1, 1, 0, 0, 0, 0));
+  y7 = _mm256_permutevar_ps(y0, _mm256_set_epi32(3, 3, 3, 3, 2, 2, 2, 2));
   y8 = _mm256_permutevar_ps(y0, _mm256_set_epi32(0, 0, 0, 0, 1, 1, 1, 1));
+  y9 = _mm256_permutevar_ps(y0, _mm256_set_epi32(2, 2, 2, 2, 3, 3, 3, 3));
 
   glmm_store256(dest[0],
                 _mm256_add_ps(_mm256_add_ps(_mm256_mul_ps(y2, y6),
-                                            _mm256_mul_ps(y4, y8)),
-                              _mm256_mul_ps(y5, y7)));
-
+                                            _mm256_mul_ps(y3, y7)),
+                              _mm256_add_ps(_mm256_mul_ps(y4, y8),
+                                            _mm256_mul_ps(y5, y9))));
 
   /* n n n n i i i i */
   /* p p p p k k k k */

include/cglm/simd/avx/mat4.h

@@ -27,8 +27,9 @@ glm_mat4_mul_avx(mat4 m1, mat4 m2, mat4 dest) {
   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 */
 
-  y4 = _mm256_permute2f128_ps(y2, y2, 0b00000011); /* d c b a h g f e */
-  y5 = _mm256_permute2f128_ps(y3, y3, 0b00000011); /* l k j i p o n m */
+  /* 0x03: 0b00000011 */
+  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 */
 
   /* f f f f a a a a */
   /* h h h h c c c c */

include/cglm/simd/sse2/mat4.h

@@ -16,7 +16,7 @@
 
 CGLM_INLINE
 void
-glm_mat4_scale_sse2(mat4 m, float s){
+glm_mat4_scale_sse2(mat4 m, float s) {
   __m128 x0;
   x0 = _mm_set1_ps(s);
 
@@ -28,7 +28,7 @@ glm_mat4_scale_sse2(mat4 m, float s){
 
 CGLM_INLINE
 void
-glm_mat4_transp_sse2(mat4 m, mat4 dest){
+glm_mat4_transp_sse2(mat4 m, mat4 dest) {
   __m128 r0, r1, r2, r3;
 
   r0 = glmm_load(m[0]);

include/cglm/types.h

@@ -26,18 +26,58 @@
 #  define CGLM_ALIGN_IF(X) /* no alignment */
 #endif
 
+#ifdef __AVX__
+#  define CGLM_ALIGN_MAT CGLM_ALIGN(32)
+#else
+#  define CGLM_ALIGN_MAT CGLM_ALIGN(16)
+#endif
+
 typedef float                   vec2[2];
 typedef CGLM_ALIGN_IF(8)  float vec3[3];
 typedef int                    ivec3[3];
 typedef CGLM_ALIGN_IF(16) float vec4[4];
 
-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 CGLM_PI    ((float)M_PI)
-#define CGLM_PI_2  ((float)M_PI_2)
-#define CGLM_PI_4  ((float)M_PI_4)
+#define GLM_E         2.71828182845904523536028747135266250   /* e           */
+#define GLM_LOG2E     1.44269504088896340735992468100189214   /* log2(e)     */
+#define GLM_LOG10E    0.434294481903251827651128918916605082  /* log10(e)    */
+#define GLM_LN2       0.693147180559945309417232121458176568  /* loge(2)     */
+#define GLM_LN10      2.30258509299404568401799145468436421   /* loge(10)    */
+#define GLM_PI        3.14159265358979323846264338327950288   /* pi          */
+#define GLM_PI_2      1.57079632679489661923132169163975144   /* pi/2        */
+#define GLM_PI_4      0.785398163397448309615660845819875721  /* pi/4        */
+#define GLM_1_PI      0.318309886183790671537767526745028724  /* 1/pi        */
+#define GLM_2_PI      0.636619772367581343075535053490057448  /* 2/pi        */
+#define GLM_2_SQRTPI  1.12837916709551257389615890312154517   /* 2/sqrt(pi)  */
+#define GLM_SQRT2     1.41421356237309504880168872420969808   /* sqrt(2)     */
+#define GLM_SQRT1_2   0.707106781186547524400844362104849039  /* 1/sqrt(2)   */
+
+#define GLM_Ef        ((float)GLM_E)
+#define GLM_LOG2Ef    ((float)GLM_LOG2E)
+#define GLM_LOG10Ef   ((float)GLM_LOG10E)
+#define GLM_LN2f      ((float)GLM_LN2)
+#define GLM_LN10f     ((float)GLM_LN10)
+#define GLM_PIf       ((float)GLM_PI)
+#define GLM_PI_2f     ((float)GLM_PI_2)
+#define GLM_PI_4f     ((float)GLM_PI_4)
+#define GLM_1_PIf     ((float)GLM_1_PI)
+#define GLM_2_PIf     ((float)GLM_2_PI)
+#define GLM_2_SQRTPIf ((float)GLM_2_SQRTPI)
+#define GLM_SQRT2f    ((float)GLM_SQRT2)
+#define GLM_SQRT1_2f  ((float)GLM_SQRT1_2)
+
+/* DEPRECATED! use GLM_PI and friends */
+#define CGLM_PI       GLM_PIf
+#define CGLM_PI_2     GLM_PI_2f
+#define CGLM_PI_4     GLM_PI_4f
 
 #endif /* cglm_types_h */

include/cglm/util.h

@@ -58,7 +58,7 @@ glm_signf(float val) {
 CGLM_INLINE
 float
 glm_rad(float deg) {
-  return deg * CGLM_PI / 180.0f;
+  return deg * GLM_PIf / 180.0f;
 }
 
 /*!
@@ -69,7 +69,7 @@ glm_rad(float deg) {
 CGLM_INLINE
 float
 glm_deg(float rad) {
-  return rad * 180.0f / CGLM_PI;
+  return rad * 180.0f / GLM_PIf;
 }
 
 /*!
@@ -80,7 +80,7 @@ glm_deg(float rad) {
 CGLM_INLINE
 void
 glm_make_rad(float *deg) {
-  *deg = *deg * CGLM_PI / 180.0f;
+  *deg = *deg * GLM_PIf / 180.0f;
 }
 
 /*!
@@ -91,7 +91,7 @@ glm_make_rad(float *deg) {
 CGLM_INLINE
 void
 glm_make_deg(float *rad) {
-  *rad = *rad * 180.0f / CGLM_PI;
+  *rad = *rad * 180.0f / GLM_PIf;
 }
 
 /*!

include/cglm/version.h

@@ -10,6 +10,6 @@
 
 #define CGLM_VERSION_MAJOR 0
 #define CGLM_VERSION_MINOR 4
-#define CGLM_VERSION_PATCH 8
+#define CGLM_VERSION_PATCH 9
 
 #endif /* cglm_version_h */

src/mat3.c

@@ -20,6 +20,12 @@ glmc_mat3_identity(mat3 mat) {
   glm_mat3_identity(mat);
 }
 
+CGLM_EXPORT
+void
+glmc_mat3_identity_array(mat3 * __restrict mat, size_t count) {
+  glm_mat3_identity_array(mat, count);
+}
+
 CGLM_EXPORT
 void
 glmc_mat3_mul(mat3 m1, mat3 m2, mat3 dest) {

src/mat4.c

@@ -26,6 +26,12 @@ glmc_mat4_identity(mat4 mat) {
   glm_mat4_identity(mat);
 }
 
+CGLM_EXPORT
+void
+glmc_mat4_identity_array(mat4 * __restrict mat, size_t count) {
+  glm_mat4_identity_array(mat, count);
+}
+
 CGLM_EXPORT
 void
 glmc_mat4_pick3(mat4 mat, mat3 dest) {

src/quat.c

@@ -8,13 +8,18 @@
 #include "../include/cglm/cglm.h"
 #include "../include/cglm/call.h"
 
-
 CGLM_EXPORT
 void
 glmc_quat_identity(versor q) {
   glm_quat_identity(q);
 }
 
+CGLM_EXPORT
+void
+glmc_quat_identity_array(versor * __restrict q, size_t count) {
+  glm_quat_identity_array(q, count);
+}
+
 CGLM_EXPORT
 void
 glmc_quat_init(versor q, float x, float y, float z, float w) {

test/src/test_affine.c

@@ -12,7 +12,7 @@ test_affine(void **state) {
   mat4 t1, t2, t3, t4, t5;
 
   /* test translate is postmultiplied */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t1, t2, t3); /* R * T */
@@ -21,16 +21,16 @@ test_affine(void **state) {
   test_assert_mat4_eq(t1, t3);
 
   /* test rotate is postmultiplied */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t2, t1, t3); /* T * R */
 
-  glm_rotate(t2, M_PI_4, GLM_YUP);
+  glm_rotate(t2, GLM_PI_4f, GLM_YUP);
   test_assert_mat4_eq(t2, t3);
 
   /* test scale is postmultiplied */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){34, 57, 36});
   glm_scale_make(t4, (vec3){3, 5, 6});
 
@@ -41,7 +41,7 @@ test_affine(void **state) {
   test_assert_mat4_eq(t3, t5);
 
   /* test translate_x */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){34, 0, 0});
 
   glmc_mat4_mul(t1, t2, t3); /* R * T */
@@ -49,7 +49,7 @@ test_affine(void **state) {
   test_assert_mat4_eq(t1, t3);
 
   /* test translate_y */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){0, 57, 0});
 
   glmc_mat4_mul(t1, t2, t3); /* R * T */
@@ -57,7 +57,7 @@ test_affine(void **state) {
   test_assert_mat4_eq(t1, t3);
 
   /* test translate_z */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){0, 0, 36});
 
   glmc_mat4_mul(t1, t2, t3); /* R * T */
@@ -65,43 +65,43 @@ test_affine(void **state) {
   test_assert_mat4_eq(t1, t3);
 
   /* test rotate_x */
-  glmc_rotate_make(t1, M_PI_4, (vec3){1, 0, 0});
+  glmc_rotate_make(t1, GLM_PI_4f, (vec3){1, 0, 0});
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t2, t1, t3); /* T * R */
 
-  glm_rotate_x(t2, M_PI_4, t2);
+  glm_rotate_x(t2, GLM_PI_4f, t2);
   test_assert_mat4_eq(t2, t3);
 
   /* test rotate_y */
-  glmc_rotate_make(t1, M_PI_4, (vec3){0, 1, 0});
+  glmc_rotate_make(t1, GLM_PI_4f, (vec3){0, 1, 0});
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t2, t1, t3); /* T * R */
 
-  glm_rotate_y(t2, M_PI_4, t2);
+  glm_rotate_y(t2, GLM_PI_4f, t2);
   test_assert_mat4_eq(t2, t3);
 
   /* test rotate_z */
-  glmc_rotate_make(t1, M_PI_4, (vec3){0, 0, 1});
+  glmc_rotate_make(t1, GLM_PI_4f, (vec3){0, 0, 1});
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t2, t1, t3); /* T * R */
 
-  glm_rotate_z(t2, M_PI_4, t2);
+  glm_rotate_z(t2, GLM_PI_4f, t2);
   test_assert_mat4_eq(t2, t3);
 
   /* test rotate */
-  glmc_rotate_make(t1, M_PI_4, (vec3){0, 0, 1});
+  glmc_rotate_make(t1, GLM_PI_4f, (vec3){0, 0, 1});
   glm_translate_make(t2, (vec3){34, 57, 36});
 
   glmc_mat4_mul(t2, t1, t3); /* T * R */
-  glmc_rotate(t2, M_PI_4, (vec3){0, 0, 1});
+  glmc_rotate(t2, GLM_PI_4f, (vec3){0, 0, 1});
 
   test_assert_mat4_eq(t3, t2);
 
   /* test scale_uni */
-  glmc_rotate_make(t1, M_PI_4, GLM_YUP);
+  glmc_rotate_make(t1, GLM_PI_4f, GLM_YUP);
   glm_translate_make(t2, (vec3){34, 57, 36});
   glm_scale_make(t4, (vec3){3, 3, 3});