diff --git a/GL/matrix.c b/GL/matrix.c
index 5c19d0a..0a06eea 100644
--- a/GL/matrix.c
+++ b/GL/matrix.c
@@ -1,7 +1,8 @@
 #include <string.h>
+#include <stdio.h>
 
 #include <dc/matrix.h>
-#include <stdio.h>
+#include <kos/string.h>
 
 #include "../include/gl.h"
 #include "../containers/stack.h"
diff --git a/GL/private.h b/GL/private.h
index 592e827..a312d28 100644
--- a/GL/private.h
+++ b/GL/private.h
@@ -98,17 +98,12 @@ typedef struct {
     GLboolean is_directional;
 } LightSource;
 
-union color_t {
-    uint8_t bgra[sizeof(uint32_t)];
-    uint32_t packed;
-};
-
 typedef struct {
     /* Same 32 byte layout as pvr_vertex_t */
     uint32_t flags;
     float xyz[3];
     float uv[2];
-    union color_t color;
+    uint8_t bgra[4];
 
     /* In the pvr_vertex_t structure, this next 4 bytes is oargb
      * but we're not using that for now, so having W here makes the code
@@ -116,14 +111,12 @@ typedef struct {
     float w;
 } Vertex;
 
-
-
 /* FIXME: SH4 has a swap.w instruction, we should leverage it here! */
-#define _SWAP32(type, x, y) \
+#define _SWAP32(x, y) \
 do { \
-    type t = *((type*) &x); \
-    *((type*) &x) = *((type*) &y); \
-    *((type*) &y) = t; \
+    uint32_t t = *((uint32_t*) &x); \
+    *((uint32_t*) &x) = *((uint32_t*) &y); \
+    *((uint32_t*) &y) = t; \
 } while(0)
 
 /*
@@ -134,16 +127,48 @@ do { \
 
 #define swapVertex(a, b)   \
 do {                 \
-    _SWAP32(uint32_t, a->flags, b->flags); \
-    _SWAP32(float, a->xyz[0], b->xyz[0]); \
-    _SWAP32(float, a->xyz[1], b->xyz[1]); \
-    _SWAP32(float, a->xyz[2], b->xyz[2]); \
-    _SWAP32(float, a->uv[0], b->uv[0]); \
-    _SWAP32(float, a->uv[1], b->uv[1]); \
-    _SWAP32(uint32_t, a->color.packed, b->color.packed); \
-    _SWAP32(float, a->w, b->w); \
+    _SWAP32(a->flags, b->flags); \
+    _SWAP32(a->xyz[0], b->xyz[0]); \
+    _SWAP32(a->xyz[1], b->xyz[1]); \
+    _SWAP32(a->xyz[2], b->xyz[2]); \
+    _SWAP32(a->uv[0], b->uv[0]); \
+    _SWAP32(a->uv[1], b->uv[1]); \
+    _SWAP32(a->bgra, b->bgra); \
+    _SWAP32(a->w, b->w); \
 } while(0)
 
+#if 0
+/* FIXME: SH4 has a swap.w instruction, we should leverage it here! */
+inline void _SWAP32( void* x, void* y) 
+{ 
+    #pragma GCC diagnostic push 
+    #pragma GCC diagnostic ignored "-Wstrict-aliasing"
+    *((uint32_t*) &x) = *((uint32_t*) &x) ^ *((uint32_t*) &y); 
+    *((uint32_t*) &y) = *((uint32_t*) &x) ^ *((uint32_t*) &y); 
+    *((uint32_t*) &x) = *((uint32_t*) &x) ^ *((uint32_t*) &y); 
+    #pragma GCC diagnostic pop 
+}
+
+/*
+    *((uint32_t*) &x) = *((uint32_t*) &x) ^ *((uint32_t*) &y); \
+    *((uint32_t*) &y) = *((uint32_t*) &x) ^ *((uint32_t*) &y); \
+    *((uint32_t*) &x) = *((uint32_t*) &x) ^ *((uint32_t*) &y); */
+
+ 
+
+#define swapVertex(a, b)   \
+{      \
+    _SWAP32(&a->flags,  &b->flags); \
+    _SWAP32(&a->xyz[0], &b->xyz[0]); \
+    _SWAP32(&a->xyz[1], &b->xyz[1]); \
+    _SWAP32(&a->xyz[2], &b->xyz[2]); \
+    _SWAP32(&a->uv[0],  &b->uv[0]); \
+    _SWAP32(&a->uv[1],  &b->uv[1]); \
+    _SWAP32(&a->bgra,   &b->bgra); \
+    _SWAP32(&a->w,      &b->w); \
+}
+#endif
+
 /* ClipVertex doesn't have room for these, so we need to parse them
  * out separately. Potentially 'w' will be housed here if we support oargb */
 typedef struct {
@@ -213,6 +238,7 @@ pvr_poly_cxt_t* _glGetPVRContext();
 GLubyte _glInitTextures();
 
 void _glUpdatePVRTextureContext(pvr_poly_cxt_t* context, GLshort textureUnit);
+void _glAllocateSpaceForMipmaps(TextureObject* active);
 
 typedef struct {
     const void* ptr;