hi,
very simple i think, but…
Does ‘VAR 2’ is only ‘fence_VAR’ ? or it is not the same way ?
Actually I use VAR on my system, but I wonder if VAR 2 could be better. I have a poor geforce 2 mx 32MO AGP DDR on a 266MHZ bus.
Thankx for replies.
VAR_2 is the same thing as VAR, except you can temporarily disable the VAR for submitting geometry not in the VAR, without forcing a flush (which would stall the pipeline; hugely expensive at times).
If you’re a good boy and keep the VAR active at all times (never disable VERTEX_ARRAY_RANGE_NV) then VAR_2 adds nothing.
At this time I’m a good boy !!!
But who knows what’s going to be tomorrow ?
When you say it’s good for submitting geometry not in the VAR, so it could be interresting if you use both VAR, CVA ?
<cough>
Why would you be using VAR and CVA together at the same time?
why ??
i don’t know now because VAR is most poweful, and is fully working…
But, maybe to draw some static objects or parts that will never change with very incessant changing data into VAR. just to separate them physically.
Even if it’s static, you want it in the VAR. You can separate parts of the VAR on your own, using NV_fence.
There’s actually very few cases where you wouldn’t want geometry in the VAR. Compiling display lists is one of the few I can think of. Even so, I currently hand the driver data not-in-VAR without disabling the VAR, and it seems to work fine (and I think I understand why that is, too, and it’s a good thing 
For geometry that arrives not-in-VAR, I actually copy into VAR and then draw; it seems faster than jiggering with disabling the VAR.
OK, but i’ve never succeed well implement the NV_fence correctly. I have not tested a lot, but… i didn’t need it in fact.
But you’re surely right.