12,396,461 members (34,199 online)
alternative version

25.5K views
9 bookmarked
Posted

# OpenGL ES: Texture2D and the Power of Two

, 19 Jul 2009 CPOL
 Rate this:

## Introduction

As most of you have figured out, OpenGL ES doesn't like textures to be non-power of two, and I really didn't want to have to force all of my textures to be powers of two before loading them into my game. This presented me with a problem that had to be solved in-code.

To really understand texture mapping in OpenGL, I had to do a lot of research. To be fair, I still don't fully understand it, although there are a few articles that do a really great job at explaining things. One of those articles is titled "OpenGL ES from the ground up" and can be located at http://iphonedevelopment.blogspot.com/2009/05/opengl-es-from-ground-up-table-of.html. Look up chapter 6 on Texture Mapping.

The problem that I was faced with, is that only powers of two can be used inside OpenGL ES, so how do you convert our texture to a power of two if it was loaded in as a 40 x 40 texture (for example)? There are several ways of doing just this, but I settled with the way Apple did it for their Texture objects, and moulded that to fit my own code. They came up with the following to convert the width and height:

```//Adjust the width and height to be a power of two.
if(( _width != 1) && ( _width & (_width - 1) )
{
i = 1;
while( (sizeToFit ? 2 * i : i) < _width)
i *= 2;
_width = i;
}

if(( _height != 1) && ( _height & (_height - 1) )
{
i = 1;
while( (sizeToFit ? 2 * i : i) < _height)
i *= 2;
_height = i;
}```

The first portion of the code will convert the width to a power of two, then the height. Since the texture will be stretched, and never shrunk, Apple had to figure out a way to shrink the texture if it became longer than the max texture size defined by `kMaxTextureSize` = 1024. The following code will do just that:

```//      scale down an image greater than the max texture size
while((_width > kMaxTextureSize) || (_height > kMaxTextureSize))
{
_width /= 2;
_height /= 2;
transform = CGAffineTransformScale(transform, 0.5, 0.5);
imageSize.x *= 0.5;
imageSize.y *= 0.5;
}```

The "`transform`" is nothing more than a `CGAffineTransform` (Core Graphics). So, now that you've glanced over the code, you remember in the article where Jeff was talking about S and T coordinates? We'll take care of those like this:

```_maxS = imageSize.x / (float)_width;
_maxT = imageSize.y / (float)_height;```

This will give us a number between 0 and 1 which can be used in our `textureCoordinates[]` for drawing. This should only render the portion of the texture that has the data we wish to make it to the screen, and not the padding we added to ensure it was a power of two.

While I am still learning OpenGL, I figured that a lot of people would be interested in the following solution... especially if you had done what I had, and thrown out the Apples `Texture2D` class while working on iPhone OpenGL applications. The full source code to the `Texture2D`, `TextureManager`, and any other class used by my C++ iPhone Engine can be found at http://code.google.com/p/djinnengine.

Hope this was informative! If anyone has any further information that would be useful, feel free to comment!

Happy coding everyone!

## History

• 19th July, 2009: Initial post

## Share

 United States
No Biography provided

## You may also be interested in...

 First Prev Next
 Why not using Texture atlas instead? Simone Serponi11-Oct-09 23:11 Simone Serponi 11-Oct-09 23:11
 Faster way to find next power of two JeanLuc_19-Jul-09 12:18 JeanLuc_ 19-Jul-09 12:18
 Re: Faster way to find next power of two Craig Giles19-Jul-09 16:47 Craig Giles 19-Jul-09 16:47
 Re: Faster way to find next power of two cmk19-Jul-09 18:33 cmk 19-Jul-09 18:33
 You can also use the _BitScanReverse() intrinsic. e.g. ```#include #pragma intrinsic(_BitScanReverse); inline unsigned long NextPow2( unsigned long I ) { _BitScanReverse(&I, I-1); return( 1 << (I+1) ); } ``` As an intrinsic _BitScanReverse will emit bsr. This is available for both x86 and x64. ...cmk The idea that I can be presented with a problem, set out to logically solve it with the tools at hand, and wind up with a program that could not be legally used because someone else followed the same logical steps some years ago and filed for a patent on it is horrifying. - John Carmack
 Re: Faster way to find next power of two yarp19-Jul-09 23:11 yarp 19-Jul-09 23:11
 Last Visit: 31-Dec-99 18:00     Last Update: 24-Jul-16 18:03 Refresh 1