Archived

This topic is now archived and is closed to further replies.

vertex compresson - too much of a good thing?

This topic is 5149 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

I have a vertex that is made up of the following: position (x,y,z) tex1 (u,v) tex2 (u,v) tex3 (u,v) I plan on compressing it and then uncompressing in a vertex shader. x and z for the position are only ever going to be integers and won''t ever be greater than 65535 so I can pack them together into 4 bytes. y will need to be a float and can be whatever I want so I will leave that alone. For the texture coordinates, I can pack each pair into 1 or 2 bytes. If I choose one then I end up with an 11 byte vertex. With 2 I end up with a 14 byte vertex. How much faster would it be to forget the compression on (xyz) and bump it up to 16 bytes? All of the developer articles I have read say that it would be best, but how much better is it?

Share this post


Link to post
Share on other sites
It may save a few cycles per vertex at transfer and transformation stages. The driver doesn''t need to pad the addresses to cache, if the vertex size is a multiple of the cache slot width.

-Nik

Share this post


Link to post
Share on other sites