Jump to content
  • Advertisement
Sign in to follow this  
piluve

DX12 Should Graphics and Compute share the same descriptor table?

This topic is 383 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

Hello!

Jumping from one bug the the next :D.

I wonder, If having the same descriptor table is a bad idea or if it should be avoided? In my rendering code, I'm only using one command list (graphics), depending if the current call is a Draw* or a Dispatch, I set the Descriptor table with SetComputeRootDescriptorTable or with SetGraphicsRootDescriptorTable.  

I'm asking this because I encountered a bug where, if I perform a draw call in between some compute, the following compute work will be all messed up and not working properly. This problem may not be related at all with the main question (I think there is something going on with my depth buffer).

Thanks!

Share this post


Link to post
Share on other sites
Advertisement

There's no rule against sharing descriptor tables across Graphics and Compute, nor do I know of any performance implications of doing that. If you had problems with the correctness of your Compute work then the chances are you missed a barrier or two.

Edited by ajmiles

Share this post


Link to post
Share on other sites

As long as you make sure that the life span of your descriptors in the table extend past the usage on the GPU, you are good to go. 

 

You may want to look at the root signature 1.1 and static versus volatile descriptor for additional driver optimisation possibilities ( basically, when the driver can assume the descriptor is ready and could be inline, that kind of thing ).

 

Edited by galop1n

Share this post


Link to post
Share on other sites
15 hours ago, ajmiles said:

There's no rule against sharing descriptor tables across Graphics and Compute, nor do I know of any performance implications of doing that. If you had problems with the correctness of your Compute work then the chances are you missed a barrier or two.

Hi, I'll double check the barriers thanks for the tip :)

4 hours ago, galop1n said:

As long as you make sure that the life span of your descriptors in the table extend past the usage on the GPU, you are good to go. 

 

You may want to look at the root signature 1.1 and static versus volatile descriptor for additional driver optimisation possibilities ( basically, when the driver can assume the descriptor is ready and could be inline, that kind of thing ).

 

Hi, that may be useful I will bookmark it. Thanks!

Share this post


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

  • 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!