glDrawRangeElements vs glDrawElements and VBOs

Probably it pre-process the range and then use it as a lookup with the indices, so there isn''t any vertex which is transformed twice.

It suppose it when i read from the manpages:

quote:

---

There is no requirement that all vertices in the range [start,end] be referenced. However, the implementation may partially process unused vertices, reducing performance from what could be achieved with an optimal index set.

---

We could suppose, that if you use 3 indices in a range of 1000 vertices, I hope the driver don''t preprocess the whole range.

_______________

Jester, studient programmer
The Jester Home in French

What if i am rendering a model and i intend to use all the vertices of the vertex array that i pass?

Should i just stick to DrawElements since i am using all the vertices in the array?



After reading through the document at nVidia''s site, i interpret the information as this:

The driver may use the range information and optimise the format of the indices array. For example it might convert 32-bit int into 16-bit int to reduce memory usage and speed up processing by a factor of 2.

Secondly it will also not need to scan the indices array, resulting in a speedup.

Is that correct?

quote:

---

Original post by jesterlecodeur
Probably it pre-process the range and then use it as a lookup with the indices, so there isn''t any vertex which is transformed twice.

---

That''s the reason I use compiled vertex arrays and glDrawElements, I don''t know if there are any benefits to using glDrawRangeElements, but I could look into it and test it in my code to see what kind of speed difference (if any) it has .

Apparently nVidia states that it is more efficient to use glDrawArrayElements even with VBOs, i don''t really know how compiled Vertex Arrays work.

But i suppose it should apply to all.

in compiled vertex arrays, you can lock and unlock the array. Locking the array tells opengl that we arent going to change anything inside of the array for now, so opengl might be able to find a way to speed it up based off this fact. When you change something, you unlock the array, which tells opengl that you are changing things. It''s pretty much like direct x''s vbo''s. The lock and unlock are available via the ARB.

I believe the locking and unlocking is stimulated in VBOs via the STATIC, DYNAMIC and the other command when defining VBOs.