• Advertisement
Sign in to follow this  

d3d11 Hardware Geometry Shader

This topic is 2478 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

In the NVIDIA "GPU Programming Guide Geforce 8 and 9 Series" they basically say that the GS is not very good, and really only recommend it for point sprites. This document is about 3 years old now, and I am just wondering if anyone has read in a newer document that they have improved things with Fermi architecture.

Share this post


Link to post
Share on other sites
Advertisement
In my experience it's generally okay as long as you don't emit too many new primitives/verts, and your post-vertex-shader vertex structure isn't too big. Using it for tessellation or to send a single primitive to multiple render targets just isn't worth it.

Share this post


Link to post
Share on other sites

With SM5 you have more powerful tool to generate new vertices/triangles - tessleation (hull & domain shaders).
For example read this: Tessellation of Displaced Subdivision Surfaces in DX11.

The problem with this is that you can only generate triangles. If you want more fine control over what you are making, the GS is a good choice. However, as MJP said, it has to be used for the right things. I have found that the sending of geometry to multiple render targets can be helpful in a number of cases, especially if there is some moderately expensive vertex level processing (i.e. skinning).

Share this post


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

  • Advertisement