In this case, is there a speed difference in clockwise or anticlockwise sorting?
- Is it right that lighting (glEnable(GL_LIGHTING)) can be very slow? (Yes, I used plane normals...)
- Has OpenGL an intern polygon counter which shows me how many polygons in a frame are drawn? When yes, how could I get this number?
I know some of this questions may be stupid.
Chris
🎉 Celebrating 25 Years of GameDev.net! 🎉
Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!
Polygon problems
Author
Here are some questions over polygons:
- How many polygones could I normally set into a scene? (e.g. P3 667 MHZ / NIVIDA RIVA TNT2 32 MB)
In this case only polygons, no other stuff (for speed testing). I''ve getting only 100000 polygons with 20 FPS on this system, is this normal?.
- Does the graphic hardware normally supports simply one colored sold polygon drawing, and do it support line drawing (I think no)?
- Which OpenGL setting have the most influence on the FPS (glEnable(GL_BLEND)...)?
- In some situations the backface culling is slower as no backface culling, I think this isn''t possible (my program fooling me)? (e.g. glEnable(GL_CULL_FACE); / glCullFace(GL_BACK)
In this case, is there a speed difference in clockwise or anticlockwise sorting?
- Is it right that lighting (glEnable(GL_LIGHTING)) can be very slow? (Yes, I used plane normals...)
- Has OpenGL an intern polygon counter which shows me how many polygons in a frame are drawn? When yes, how could I get this number?
I know some of this questions may be stupid.
Chris
In this case, is there a speed difference in clockwise or anticlockwise sorting?
- Is it right that lighting (glEnable(GL_LIGHTING)) can be very slow? (Yes, I used plane normals...)
- Has OpenGL an intern polygon counter which shows me how many polygons in a frame are drawn? When yes, how could I get this number?
I know some of this questions may be stupid.
Chris
how do you display the 100000 polygons?
with glVertex() or glDrawElements()?
or do you use a displaylist?
// Just wondering, Ankan
with glVertex() or glDrawElements()?
or do you use a displaylist?
// Just wondering, Ankan
i just get around 50 fps when drawing 200 untextured and unlit triangles.... and you get 20 fps when drawing 100000, you should be happy!!!, I have a celeron 550 Mhz with a matrox G400 with latest drivers. What can be wrong with my code if i''m only getting 50 fps @ 200 triangles? i''ve taken lesson 1 and just made a loop which iterates 200 times and 3 glVertex() calls each loop. What''s wrong with that?
Ankan
Ankan
The slowness probably comes from this:
You are making three calls to glVertex3f() each iteration, why not use a vertex array and make one call to glVertex3fv() one call, and only one parameter.
Or, you might try putting the object in a display list.
Try something like that.
-Mezz
You are making three calls to glVertex3f() each iteration, why not use a vertex array and make one call to glVertex3fv() one call, and only one parameter.
Or, you might try putting the object in a display list.
Try something like that.
-Mezz
When a company says their hardware can display 20 millions polygons per second (like the GeForce), it is only a best case scenario.
These are only flat-shaded, no lightning, no whatever, triangles drawn with one vertex definition per triangle (fan-strips, for example, which use, I think, 2+n vertex for n triangles).
Thus, on this best case scenario, you won''t have more than 50FPS for 400k triangles.
You must not forget the fill-rate either, but I don''t think it will make any noticeable difference unless you use high resolutions or FSAA.
EL
These are only flat-shaded, no lightning, no whatever, triangles drawn with one vertex definition per triangle (fan-strips, for example, which use, I think, 2+n vertex for n triangles).
Thus, on this best case scenario, you won''t have more than 50FPS for 400k triangles.
You must not forget the fill-rate either, but I don''t think it will make any noticeable difference unless you use high resolutions or FSAA.
EL
- In some situations the backface culling is slower as no backface culling, I think this isn''t possible (my program fooling me)? (e.g. glEnable(GL_CULL_FACE); / glCullFace(GL_BACK)
In this case, is there a speed difference in clockwise or anticlockwise sorting?
don''t think so
- Is it right that lighting (glEnable(GL_LIGHTING)) can be very slow? (Yes, I used plane normals...)
one directional light is on nvidia processors free
- Has OpenGL an intern polygon counter which shows me how many polygons in a frame are drawn? When yes, how could I get this number?
nope u have to count them yourself
You have to have something wrong with your card man. I have a 2/ppro233 Matrox Millenium and I get about 20 fps with around 2.5k lit, nontextured triangles. Thats with one directional and one positional spot light. Use display lists and vertex arrays when ever possible. I usually get a 10 fps jump using one of the two over immediate mode. Don''t forget backface culling.
ankan666 make your window smaller + watch the fps rise, the reason being glClear(..) isn''t free.
youre getting 50fps with 200 polys try 2000polys u should get about 45fps
youre getting 50fps with 200 polys try 2000polys u should get about 45fps
Good point about glClear(...). If you're inside a totally enclosed draw area (like inside a skybox/dome or whatever), then just skip clearing the color(frame) buffer each frame. Even if you're not enclosed, try it... it looks way trippy
You still need to clear the depth buffer though (if you use it)... there is trick that you can use to avoid this as well, but I can never remember the damn thing
More generally speaking, something to avoid in OpenGL is state changes. Switching features on and off willy-nilly will see frame rates plummet, especially changing texture state (swapping texture objects).
Also, avoid redundant vertices whenever possible. Do you really need to specify 3 vertices for every triangle, or are some shared? And try to pack as much stuff between each glBegin(..) and glEnd() as you can.
ahh, random thoughts of a disturbed man...
-------------
squirrels are a remarkable source of protein...
Edited by - Bad Monkey on July 11, 2000 2:37:18 AM
You still need to clear the depth buffer though (if you use it)... there is trick that you can use to avoid this as well, but I can never remember the damn thing
More generally speaking, something to avoid in OpenGL is state changes. Switching features on and off willy-nilly will see frame rates plummet, especially changing texture state (swapping texture objects).
Also, avoid redundant vertices whenever possible. Do you really need to specify 3 vertices for every triangle, or are some shared? And try to pack as much stuff between each glBegin(..) and glEnd() as you can.
ahh, random thoughts of a disturbed man...
-------------
squirrels are a remarkable source of protein...
Edited by - Bad Monkey on July 11, 2000 2:37:18 AM
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
