PDA

View Full Version : One big poly or 4 smaller ones?



padlamoij
01-14-2003, 10:06 AM
I don't know a lot about hardware so this may be really obvious but...
would it be faster to do one big triangle strip square with a 1024 x 1024 texture on it or 4 smaller triangle strip squares with 512 x 512 textures on them.
And yes, I need the resolution to be that high.
I am using for a 2D background, and if anyone knows a better way to do a 2D background in openGL i'm all ears...literally i'm made entirely of ears. I hear a lot of stuff. Okay sorry i'm done. Thank you.

mdog1234
01-14-2003, 10:40 AM
I would think it would be faster to do a a large one. I would just try it both ways and see which way is faster for you. If you have a newer video card I would bet it doesn't really make much of a difference either way

padlamoij
01-14-2003, 10:45 AM
ya, i figured anything past a geForce 2 shouldn't mind a texture that big but like i said i'm far from a hardware guru. Or any guru for that matter. I'll probably just leave it as one big one for now. If i decide to try out splitting it up later i'll post the results. Thanks.

HamsterofDeath
01-16-2003, 02:03 AM
for the smaller ones, u need 4 glbindtextures -> this will make it definetly slower

padlamoij
01-16-2003, 10:19 AM
good point. I hadn't even thought about that.
but all that will be done during non-time-critical processing. I was more worried about the big fat texture slowing things down during the rendering. Of course i don't know if a bigger texture would have anything to do with rendering speed after it has been bound to the polys. Like i said i'm not a guru of any sort. thanks for the info.

oliii
01-18-2003, 04:27 AM
You can blit a bitmap to the screen. Can't remember the exact command, but you have to specify pixel packing and stuff. Also the bitmap has to be the same resolution as the display. Not sure if it is faster. Probably not.

The advantage of doing texture mapping is that you can change the resolution of the display. Good for multiplatform support and user settings.


[This message has been edited by oliii (edited 01-18-2003).]

chxfryer
01-18-2003, 09:12 AM
why would you need 4 texture binds. Just bind the whole bitmap once and change the texture coords as needed for the 4 quads.

oliii
01-18-2003, 09:29 AM
The advantage to use 4 quads would be to make sure it works if the hardware can't handle massive textures. Breaking the bitmap down to smaller cells would help. A big poly is faster than 4 polys, yeah. Appart from that, ....

padlamoij
01-18-2003, 11:31 AM
ya that was my concern...the big texture. I don't know a lot about the hardware of graphics cards. I figure that new cards shouldn't have a problem with it but i was wondering if the older cards like around the geforce 2 level would. Also i was wondering if it would be faster or slower for all cards in general if i split up the texture. Point taken about splitting the texture and not the polys. Maybe i will just have to find the time to try out this experiment.

M/\dm/\n
01-18-2003, 11:52 AM
You can write routine, where program takes 2 frames, 1st with 1 poly, 2nd with 4/4+ polies & calculates time, and then uses fastest method.
BTW, I have similiar problem, my skybox has quite good textures, but I had to make them lower res to keep it realtime like, now it looks like crap. They WERE attached to one polly each. Currently I'm rewritting my main.cpp http://www.opengl.org/discussion_boards/ubb/mad.gif so I have no time to check out the best method, so I'd like to hear what to do, if you'll find out.