Is MultiDrawArrays significantly faster than call DrawArrays multiple times?
Is it worth to gather the DrawArrays’ data to pointer arrays and then call a single MultiDrawArray?
Has anyone any test from this?


FWIW, I only experienced MultiDrawArrays on ATI R9700 which was (is?) painfully slow when rendering too short triangle strips.

I did not notice any performance gain by using MultiDrawArrays - but it may have changed is lastest drivers, and may not be true for all platforms.

The best thing to do is to test, and let us know what you found !


Originally posted by Nicolas Lelong:

The best thing to do is to test, and let us know what you found !

I have test it. There is no performance increase/decrease on my machine:
AthlonXP1800+, GeforceTi4200, I have used VBO in the test.

I’ve tested that across the entire Radeon range (R100, R200, R300) and on a Geforce3 and a Geforce2MX. Nothing.

Performs exactly the same as the ‘official’ replacement code (using multiple glDrawArrays). But then, my geometry transfer is not optimal.

It might have been worthwhile to keep supporting it in anticipation of future driver/hardware improvements, but the additional management overhead wasn’t worth it to me.