PDA

View Full Version : Typical 3D accelerators: integer or floating-point...which most efficient?



08-08-2001, 06:00 PM
As far as maximum performance (i.e. in a game) is concerned, what will yield the best results on a typical 3D accelerator using OpenGL: the integer or floating-point versions of such routines as glVertex, glColor, etc.? Thanks!

Akash
08-08-2001, 06:45 PM
To be honest, i am not sure, but i think integer ones are suppose to be faster. But on the other hand floats are more precise then ints. So even though integer ones will run a bit faster it wont be as precise as ints. For example 0.4 will be interpreted as 0 and 0.5 as 1.

hope it helps

[This message has been edited by Akash (edited 08-08-2001).]

Zadkiel
08-09-2001, 08:15 AM
I read on this forum once that OGl uses floats internally.

If thats the case then floats will be the fastest as there won't have to be any conversions.

08-09-2001, 08:40 AM
since OpenGL uses floats internally and uses floats for pin-point accuracy, you should DEFINITELY use floating points. it will be very accurate, and haveing to convert to integers at every turn will turn PAC-MAN into PAC-SNAIL. so use floats for all graphics commands, and ints for anything else (ie bullets, lives, etc)

08-09-2001, 11:38 AM
Thanks a bunch. Althought I've been using OpenGL for quite a while, I was unsure on that one. Thanks!