• Advertisement
Sign in to follow this  

DirectX9 depth map rendering

This topic is 3315 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, i read this post and wanted to reply in it but it is expired: http://www.gamedev.net/community/forums/topic.asp?topic_id=401503&forum_id=10&gforum_id=0 so i am going to try my question this way. Is the performance boost significant enough to implement this sort of hacky way of rendering depth map next to the usual directx9 way (G32R32F format for example). Also i was wondering (i tried googling) if there were any further general tips on rendering a depth map the usual way. I somehow have the feeling that right now this part of the engine is far to slow comparing to what i expected (i get 100 FPS when generating 4 shadow maps on 8800Ultra, with the scene itself also, rendering about 13 000 triangles in total). I right now disable the B and A color channels when writing to the G32R32F render target, also i write to a texture atlas. Does anyone have any information on any of these two questions?

Share this post


Link to post
Share on other sites
Advertisement
Are you sure the bottleneck is the fillrate? DirectX 9 is known for its DrawPrimitve bottleneck. So if you render the scene multiple times, each time using many draw calls, your application will become CPU limited very quickly.
In that case, having a super high end GPU does not make any difference. So, install nvidia performance hud and inspect the rendering bottleneck, before assuming its about the number of triangles or the screen size or what ever.

Greatz,
Dietger

Share this post


Link to post
Share on other sites
i tracked down a lot with nvidia PerfHUD already. I am not experienced yet (started working with it 2 days ago) but the numbers give me the impression my card should be able to do lots more. What are generally good things to look at in perfHUD?

and i am talking about rendering a depth map, and yes this is for shadowing :)

Share this post


Link to post
Share on other sites
Nvidia GPU's and newer ATI GPU's will rener 2x the pixels when you render depth only...this is an advantage if you're fillrate-bound however for shadowmaps it's very common to be limited by triangle setup.

However the real advantage of hose techniques is that you can use less texture samples in the pixel shader, which can help a lot when you're using large PCF kernels.

Share this post


Link to post
Share on other sites
well i think the problem might not be the triangle amount but more the state changes. I read about switching of SaveState etc at the effect.Begin(FX) part. I can maybe (if pix allows me) upload a file with all the calls one frame does... :)

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement