iphone - PVR textures versus PNG in OpenGL ES -


I am developing a 2D application for the iPhone which renders a lot of textures, most of them from PNG files At the moment, alpha is loaded with transparency. POV-textures are loaded with PVR-textures that come with a crash landing example. PVR-testers are loaded from the PVR detection class by PVR extraction loader example. I make PVR textures using APL texture tools.

As a test, I render a background (512 * 512) and 36 90 * 64 pixel prites (512 * 512 texture) with transparency. The PVR texture is at approximately 58 fps and PNG 47 fps. Can I expect this or what difference should be big? In addition, the text generated by TexturTol really looks bad, is PVRTEST better?

PNG:

  • High precision color representation, not damaged
  • Slow decryption from disk.
  • Slow uploading on graphics hardware is done by internal pixel sorting drivers.
  • Because of the limited memory bandwidth, the actual amount of slow rendering slump depends on the scenario.
  • Take more RAM / VRAM space.
  • Editable, filter / mixed / size can be changed / converted before uploading your software.
  • MIP-Maps can be automatically generated during texture upload
  • Using disk space varies with content, for very simple images, almost photorealistic people Uncompressed for
  • Any image-editing software can be exported directly and quickly. PVR:
    • Less precise lossy compression 2 compression levels, 2 bit per pixel and 4 bit per pixel are available. Blocky can damage sharp edges and smooth gradient. Image quality changes with 3 or 4 color channels, so that you can use alpha channels, but lossy compression can produce undesirable results.
    • Fast loading from the disk, no software decompression is necessary.
    • Approximate immediate texture upload, because it is an internal hardware format, which will change between drivers.
    • Due to the fast small memory bandwidth usage rendering pixel rendering speed is mostly limited by other factors, when PVR textures are used.
    • In a small amount of RAM and amp; VRAM Space
    • MIP-Maps should be pre-generated.
    • You can not generate or edit PVR inside your software AFAIK or it will be very slow.
    • The use of disk space can be compressed further than other methods directly proportional to the image size.
    • Size Limits
    • An additional conversion tool is required, the process can be automated, but the build time will be significantly slowed.

Comments