[Qt-interest] Performance issues with Graphics View Framework

[Constantin Makshin]

On Tue, 21 Jul 2009 14:15:52 +0400, Enrico Ros <enrico.qt at email.it> wrote:
>> If you want to use available hardware acceleration, then do what Enrico
>> suggests (load the image as a texture and scroll it by modifying texture
>> coordinates). Updating 4 vertices is MUCH cheaper than uploading new  
>> image
>> to VRAM (assuming each vertex has 3 coordinates (x, y, z) and 2 texture
>> coordinates (s, t), 4 vertices'll occupy only 80 bytes).
>
> Yes, and he could go even further, by using a vertexbuffer with the data  
> and just scrolling the texture matrix :D 1 operation ~ 16 bytes ;-)  
> Impressive savings.
OpenGL uses 4x4 matrices, so it would be 64 bytes. Still less than 80. :)

Simone can go even further than you suggest and have even more impressive  
savings by using a simple vertex shader and passing only 1 (one!) float to  
it. :D

But I think changing texture coordinates directly is enough for Simone's  
needs.

-- 
Constantin "Dinosaur" Makshin