Does Interleaved Vertix Attributes Arrays really Helps Performance
Reading the following two articles,
http://www.opengl.org/wiki/VBO#Inner_Workings
http://www.sci.utah.edu/~bavoil/opengl/bavoil_trimeshes_2005.pdf
I can't convence myself on a conclusion of using Interleaved Vertix arrays to improve Cache performance.. and the reason is that if you want to use glDrawArrays(),which is considered slow as compared to glDrawElements, then you are duplicating alot of vertices (a waste of GPU memory) if you want to use glDrawElements() than you don't have spatial locallity in your Vertix attributes buffer.... since you are refering to them with Indices from your IBO.
So my question is. Is it really worth it? creating an Interleaved Attributes arrays and a single VBO
Can some OpenGL guru Guide me here?
Quote:Original post by AverageJoeSSU
http://developer.nvidia.com/object/bindless_graphics.html
Thats koooooooool.. but its not a solution to my question. its rather an alternative.
Quote:Original post by waZimNot necessarily true. Correctly cache-optimized vertex buffers will have reasonable spatial locality. In any case, it's not a huge thing to worry about. Readahead on an index buffer is trivial, so memory access latency is hidden.
if you want to use glDrawElements() than you don't have spatial locallity in your Vertix attributes buffer
Quote:So my question is. Is it really worth it? creating an Interleaved Attributes arrays and a single VBOWell, it's easy to implement. Try it and find out.
from my testing IIRC I saw about 3% improvement (this was in a benchmark) for a real scene u will prolly notice no difference
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement