TextureAtlas Inconsistency?

[Gary Daniels]

I noticed that if I create a TextureAtlas with a power of 2 size
(1024x1024), the tex_coords of a 16x16 image added are:

(0.0, 0.0, 0.0, 0.015625, 0.0, 0.0, 0.015625, 0.015625, 0.0, 0.0,
0.015625, 0.0)

With a size of 1023x1023 (or any non-power of 2), they are pixel
based:

(0.0, 0.0, 0.0, 16.0, 0.0, 0.0, 16.0, 16.0, 0.0, 0.0, 16.0, 0.0)

Is this intentional?

[Nathan]

Gary,

I don't have any previous experience with that portion of pyglet, but
looking at the source code for TextureAtlas (image/atlas.py:208 from
hg tip), it appears that it intends to return a TextureRegion with the
same pixel-dimensions as the original image.

So my guess would be that that's a bug.

That seems like a pretty big image to be making an atlas of. Are you
sure this behavior happens with _any_ image with a size that's a power
of 2? (e.g. 64x64?)

~ Nathan

> --
> You received this message because you are subscribed to the Google Groups "pyglet-users" group.
> To post to this group, send email to pyglet...@googlegroups.com.
> To unsubscribe from this group, send email to pyglet-users...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/pyglet-users?hl=en.
>
>

[Gary Daniels]

On Nov 19, 11:48 pm, Nathan <nathan.sto...@gmail.com> wrote:
> Gary,
>
> I don't have any previous experience with that portion of pyglet, but
> looking at the source code for TextureAtlas (image/atlas.py:208 from
> hg tip), it appears that it intends to return a TextureRegion with the
> same pixel-dimensions as the original image.
>
> So my guess would be that that's a bug.

Yeah, I saw where it doing it, I just wanted to know if I could count
on that behavior or if it was a bug that would eventually be fixed.
I've written a C extension to blit the tiles and that code needs it to
be one way or the other consistently. For now, I can control it by the
size of the atlas.

> That seems like a pretty big image to be making an atlas of.

I am using the atlas to store a currently unknown number of 16x16
tiles. I just picked 1024x1024 as a starting point.

>Are you sure this behavior happens with _any_ image with a size that's a power
> of 2? (e.g. 64x64?)

So far I've only added 16x16 and 256x16 images, but they are both one
way or the other depending on the power of 2-ness of the atlas.

Thanks

[Andreas Schiefer]

I'm not quite sure, but the implementation may use the ARB_texture_rectangle extension for non-power of 2 textures. This extension specifies texture coordinates to be pixel based/non-normalized.

So I guess the pixel based texture coordinates for non-power of 2 textures are probably intentional (to conform with the extension).

Note that such textures do not support mipmap filtering (if you need that).

Andy

[Gary Daniels]

On Nov 20, 3:29 am, Andreas Schiefer <andreas.schie...@gmail.com>
wrote:

> I'm not quite sure, but the implementation may use the
> ARB_texture_rectangle extension for non-power of 2 textures. This extension
> specifies texture coordinates to be pixel based/non-normalized.
> So I guess the pixel based texture coordinates for non-power of 2 textures
> are probably intentional (to conform with the extension).
> Note that such textures do not support mipmap filtering (if you need that).

That is what it is doing.

Do you have any idea why this works:

...
self.tileAtlas = pyglet.image.atlas.TextureAtlas(1025,1025) #non po2
... add some 16x16 images

glEnable(self.tileAtlas.texture.target)
glBindTexture(self.tileAtlas.texture.target,
self.tileAtlas.texture.id)
glBegin(GL_QUADS)
glTexCoord2i(0,0)
glVertex3i(0,0,0)
glTexCoord2i(1025,0)
glVertex3i(1025,0,0)
glTexCoord2i(1025,1025)
glVertex3i(1025,1025,0)
glTexCoord2i(0,1025)
glVertex3i(0,1025,0)
glEnd()

and displays the whole texture on the screen.

But this does not work:

...
self.tileAtlas = pyglet.image.atlas.TextureAtlas(1024,1024) #po2
... add some 16x16 images

glEnable(self.tileAtlas.texture.target)
glBindTexture(self.tileAtlas.texture.target,
self.tileAtlas.texture.id)
glBegin(GL_QUADS)
glTexCoord2f(0.0, 0.0)
glVertex3f(0,0.0, 0.0)
glTexCoord2f(1.0, 0.0)
glVertex3f(1.0, 0.0, 0.0)
glTexCoord2f(1.0 ,1.0)
glVertex3f(1.0, 1.0, 0.0)
glTexCoord2f(0.0, 1.0)
glVertex3f(0.0, 1.0, 0.0)
glEnd()

If I add:

glMatrixMode(GL_PROJECTION);
glPushMatrix();
glLoadIdentity();

before the glBegin, it does display the texture, but almost freezes my
video and it gets about .01 fps.

[Gary Daniels]

> Do you have any idea why this works:
>

Nevermind. I was assuming since I was calling glVertex3f instead of
glVertex3i that I had to use 0.0 - 1.0 instead of pixel coords. This
works:

glEnable(self.tileAtlas.texture.target)
glBindTexture(self.tileAtlas.texture.target,
self.tileAtlas.texture.id)
glBegin(GL_QUADS)
glTexCoord2f(0.0, 0.0)

glVertex3f(0.0,0.0, 0.0)
glTexCoord2f(1.0, 0.0)
glVertex3f(1024.0, 0.0, 0.0)
glTexCoord2f(1.0 ,1.0)
glVertex3f(1024.0, 1024.0, 0.0)
glTexCoord2f(0.0, 1.0)
glVertex3f(0.0, 1024.0, 0.0)
glEnd()