arrays of floats or array of structures for opengl ?

Hello,

I am "researching" vertex buffer objects...

I am wondering how to combine positions with texture coordinates.

It seems there are two possibilities:

array of position;
array of texture_coordinate;

or

array of structure;

Which this document shows:

http://www.opengl.org/wiki/Common_Mistakes

The opengl doc mentions "stride".

Confusion would be with the possibility of memory being in system memory or
video memory... the programmer don't know...

Maybe the graphic cards has an internal format for verteces ? Does it use
arrays or does it use structures ?

In other words:

What would be the best memory layout for performance ?

Strides would add gaps... so I am thinking... maybe array of floats is
better ?!?

This is one of those things that must be decided upon perform starting to
code a lot of stuff... because it would require significant code changes in
the future if one has to switch.

For now I am still in experimenting phase

Though even for me doing a benchmark would require extra programming time ?!

Bye,
Skybuck.

This document is nice:

http://www.opengl.org/resources/code...01/perfogl.pdf

It mentions something interesting:

"Depth testing is expensive..."

I am not sure if this includes verteces going out of view or so... in 2d
mode...

Maybe depth testing is not required to make verteces go out of view by the z
value as I was planning to do...

Actually this is probably not an issue at all anymore... since each verteces
will be determined after each round or so ?!

But not completely sure.

So interesting read anyway.

Bye,
Skybuck.

This sux:

"
Once again, for certain architectures, this may be the best format for
specifying geometry. The only way
to know is to conduct a benchmark case
"

Would be nice if opengl/drivers had some hints for the programmer for a
change !

Bye,
Skybuck