clipping: simplify texture coordinate transformation

Instead of scaling the coordinates by -1 on the Y axis, modify the
texture data by flipping it vertically. This allows us to use
guMtxTransApply() function, which is way faster than guMtxConcat().

src/clip.c

@@ -41,10 +41,13 @@ POSSIBILITY OF SUCH DAMAGE.
 
 static GXTexObj s_clip_texture;
 static uint8_t s_clip_texels[32] ATTRIBUTE_ALIGN(32) = {
- 0x0f, 0x00, 0x00, 0x00,
+ /* We only are about the top-left 2x2 corner, that is (given that pixels
+ * are 4 bits wide) the first and the fourth byte only.
+ * Note how the positive pixel value is set on the bottom right corner,
+ * since in OpenGL the y coordinate grows upwards, but the t texture
+ * coordinate grows downwards. */
  0x00, 0x00, 0x00, 0x00,
- /* There are six more rows, but we don't care about their contents: we only
- * care about the top-left 2x2 corner */
+ 0x0f, 0x00, 0x00, 0x00,
 };
 
 static void load_clip_texture(u8 tex_map)
@@ -103,12 +106,7 @@ static bool setup_tev(int *stages, int *tex_coords, int tex_maps, int *tex_mtxs,
  * centre of our texture is (0.5, 0.5), therefore we need to map the zero
  * point to that. We do that by translating the texture coordinates by 0.5.
  */
- const static Mtx trans = {
- { 1, 0, 0, 0.5},
- { 0, -1, 0, 0.5},
- { 0, 0, 0, 1}, /* row is unused */
- };
- guMtxConcat(trans, m, m);
+ guMtxTransApply(m, m, 0.5, 0.5, 0);
  GX_LoadTexMtxImm(m, tex_mtx, GX_MTX2x4);
 
  GX_SetTexCoordGen(tex_coord, GX_TG_MTX2x4, GX_TG_POS, tex_mtx);