Jump to content

  • Log In with Google      Sign In   
  • Create Account

performance question


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
3 replies to this topic

#1 lomateron   Members   -  Reputation: 342

Like
0Likes
Like

Posted 15 February 2013 - 10:00 AM

I am rendering to a texture, I am rendering many intances of a vertice that will have different locations in the texture I am going to render to, but then what happens when more than 1 vertice points to the same pixel of a texture, and I have blend enabled, does it loses all the pixel shader parallel performance?



Sponsor:

#2 Seabolt   Members   -  Reputation: 633

Like
0Likes
Like

Posted 15 February 2013 - 10:48 PM

Not really, since which ever vertex gets interpolated second will have to run after the other pixels in order to blend properly and can't really run in parallel.


Perception is when one imagination clashes with another

#3 EWClay   Members   -  Reputation: 659

Like
0Likes
Like

Posted 16 February 2013 - 09:50 AM

Short answer: no. GPUs have many specialised units for different tasks, and even if one part of the process (in this case, the final blended output of multiple quads at the same location) cannot be parallelised, the rest still will be.

#4 Jason Z   Crossbones+   -  Reputation: 5045

Like
0Likes
Like

Posted 16 February 2013 - 11:37 AM

I am rendering to a texture, I am rendering many intances of a vertice that will have different locations in the texture I am going to render to, but then what happens when more than 1 vertice points to the same pixel of a texture, and I have blend enabled, does it loses all the pixel shader parallel performance?

 

No, not really.  Since there are many shader units, the pixels can be processed in parallel, then passed on to the output merger where they are blended.  This is more or less the same situation as when the pixels aren't overlapping - the OM has to serialize the writing to a particular portion of the render target.






Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS