• Advertisement
Sign in to follow this  

HLSL - Domain / Geometry Question

This topic is 507 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 All,

 

I have a question, at the moment I am creating tessellated surfaces quite happily in my ds/hs stages.  My question is, that I am also doing my gerstner wave generation in this stage as well.

 

Would I benefit from using the Geometry stage to do the wave calculations and matrix multiplication.  Is there any performance impact by using HS/DS over GS for vertex specific transformation?

 

It's a very open ended question.  I know, but i'm interested to know whether the pipeline is improved by spreading the load more across the stages.

 

Thoughts?

 

Thanks

Share this post


Link to post
Share on other sites
Advertisement

In my experience, I've usually found that you're better off using less shader stages if you can. At a minimum using a GS will add more threads that need to be executed, and for the GS in particular there's the potential that simply using it at all will cause the hardware to have to do things that have low performance (like spilling out all of the per-vertex data to off-chip memory). On some hardware you can avoid most of these issues by having a "pass-through" GS that is always 1:1 with input and output primitives (in other words, it doesn't remove or add primitives). But that's not a guarantee, and the only way to know for sure is to profile.

Share this post


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

  • Advertisement