Adjusting the data size

When you use the correct data sizes for textures and shaders you reduce the memory requirements. Adjust the data size to reflect the requirements of your application:

If the screen resolution of the target devices is low, adjust the resolution of the textures to match that of the target screen. You can combine image downscaling with image compression. Try to find a compromise between required visual quality and application performance. See Compressing textures.
The memory required by a texture corresponds directly to the texture format you use. For example, an 8-bit grayscale texture requires 8 bits per pixel, whereas RGBA8 requires four 8-bit channels.
Make sure that the image and texture formats match. When image and texture formats do not match, Kanzi Engine converts them when your during the loading of your Kanzi application using the CPU of the target platform. This can significantly increase the amount of time it takes to launch your application.

For example, if a texture is grayscale and uses transparency, in Kanzi Studio set the texture Format property to Grayscale alpha (8-bit) instead of, for example, RGBA (8-bit). A 256 by 256 pixels RGBA texture uses 262 kb (256 * 256 * 4). When you use grayscale with alpha channel, the same texture uses 131 kb (256 * 256 * 2).

When image and texture formats do not match, Kanzi Studio prints this warning to the Log window:
```
Application WARNING: The format of the texture '<TextureName>' does not match the format of the image '<ImageName>'.
	To improve the performance of your application, make sure the formats of the image and the texture match.
```
Make sure that all faces of a cubemap texture use images of the same size and format. When size and format of the images in a cubemap texture do not match, Kanzi Engine converts them during the loading of your Kanzi application using the CPU of the target platform. This can significantly increase the amount of time it takes to launch your application.

For example, use for all faces of the cubemap texture images that are 256 by 256 pixels large and are 8-bit grayscale.

When the size and format of images in a cubemap texture do not match, Kanzi Studio prints these warnings to the Log window:
```
Application WARNING: The faces of the cubemap texture '<TextureName>' use images of different sizes.
	To improve the performance of your application, make sure all faces use images of the same size.
Application WARNING: The faces of the cubemap texture '<TextureName>' use images with different data formats.
	To improve the performance of your application, make sure the images of all faces use the same data format.
```
Use mipmaps to create a set of downscaled sub-levels from a large texture. Mipmaps double the GPU memory use, but improve the peformance when full texture does not have to be sampled (for example, when a texture is far and covers only a few pixels). See Using mipmaps.
Avoid creating textures from very large images (for example, digital camera pictures). Instead create textures from preprocessed and properly detailed graphics.
Set vertex data types from the default float type to the smaller half-float. See Setting the correct size of vertex cache.
Make sure the tools you use for creating and converting images create images of the smallest possible size.

Adjusting the data size

See also