Jump to content
  • Advertisement

Archived

This topic is now archived and is closed to further replies.

BlueChip

frustum + geomipmapping < fillrate

This topic is 5276 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi... many times I have listened things about fillrate limitation... yesterday, my terrain had 524288 triangles... ( 16 chunks of 129*129 vertex rendered with brutal force ) today, with frustum and geomipmapping my terrain has 4000-10000 triangles ( very very less ) ... but my FPS is gone from 35-40 to 50-55 ( these are results in critic zones ... in other areas the frame rate flies over 90 FPS ) I render each chunk 4 times... Is this the ill-famed fillrate limitation? or in your opinion there are some dark points? I hoped of having a gain of 40-50 frames... with a total of 75-80 FPS ( in haevy areas ) My fear, is that with a sky system + furnishing + NPG + AI + particle system, my fps become too much small

Share this post


Link to post
Share on other sites
Advertisement
quote:

The best thing to do is profile your code



sorry but what do mean "profile"?
Babylon doesn''t help me....

Share this post


Link to post
Share on other sites
A profiler tells you how much time your code spends in each of it''s functions. For instance, if it says your culling code is taking 40% of your CPU time odds are you could improve your culling routine and get a vast improvement. Without a profile you''re just kinda guessing where your bottleneck is.

If you want (and your code allows), keep your culled and mipped data. Render this over and over (ie: don''t keep running the culling and morphing code). How''s the speed? If it''s still low you can infer that rendering is a problem. Whether because of fill rate, excessive state and texture changes, improper buffer usage, etc, we don''t know that yet, but we''ve narrowed it down to your renderer only.

If you can''t find a profiler (some are free, so that''d be hard... there''s one built into VC6.0, but not .NET), tests like these may help you narrow it down.

Also you can wrap major parts of your engine in QueryPerformanceCounter calls, which is almost as good as a profiler.

QPC
Get input
QPC
update world
QPC
Cull
QPC
Geomip
QPC
Render
QPC
Apply effects
QPC
Present
QPC

You can then use these numbers to see where your time goes. By taking the QPC (QueryPerformanceCounter) result before and after the render call you see exactly how long your code remains in that function.

Other classic tests. Think you''re fillrate limited? Reduce the screen resolution of viewport size. Did your framerate change accordingly? If yes, you probably are fillrate limited. Reduce your texture sizes... Did your framerate jump up? Perhaps you''re memory limited. Both nVidia and ATI have bottleneck finding flowcharts in their developer sections (possibly hidden inside GDC whitepapers, which you should be reading anyway).

Share this post


Link to post
Share on other sites
If this is the first thing you have done on the engine, I''d just say don''t worry about it. Too many times I myself have gotten bogged down worring about the framerate of my terrain...but in all truth, no matter how many polygons you have (to an extent) the terrain will never give you a very high level of performance. Sure you can get really neat effects and stuff but rendering however many million pixels of data per second just can''t get any faster.

You can try reducing your texture detail to compensate for this. I found that no matter how many polys were in my terrain (I reached the max for 16-bit indices and there was no difference) it was the fact that the texture was wrapping so many times that cut my FPS from a solid 60 to a solid 30. It just quit rendering at the highest speed because it couldn''t do all of the calculations.

To sum, I wouldn''t worry about the speed of the terrain. Get the rest of your game done first; rendering 10 chairs, a few houses and doing AI calculations won''t hurt as much as you think.

Share this post


Link to post
Share on other sites

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!