Project: | YafaRay |
Component: | YafaRay Core |
Category: | feature request |
Priority: | normal |
Assigned: | David Bluecame |
Status: | closed |
Hi there
I wonder if the algorithm we have to subdivide final tiles to take advantage of available threads could be implemented in border rendering, as it is a feature I use very ofter for tests phase. Sometimes I even use border rendering with a 500% render scale for testing effects on small areas, but it would be great if the tiles rendered can take advantage of that nice algorithm we have implemented for speeding up full renders. Greetings
Comments
#1
Hello,
The tile subdivision algorithm works independently of rendering a full camera view or a border rendering, so you should not have any issues.
What the algorithm does is this:
//If blocksize is more than 4, resubdivide the last (2 x nunber of threads) so their block size is progressively smaller (better CPU/thread usage in the last tiles to avoid/mitigate having one big tile at the end with only 1 CPU thread)
So the only aspects that influence the algorithm are the initial tile size, the number of threads and you have to keep in mind that the minimum tile size is 4x4 (it will not automatically subdivide further) because I could see that performance degrades too much for tiles smaller than 4x4
However if you have problems with border rendering please provide an example scene and a description of the problems you are experiencing.
I hope this helps.
#2
Indeed it works, sorry
My yaf work ·Grey18 workflow ·Sampling strategy ·SPPM explained ·[url=h
#3
No problem, I wish all problems were solved like this ;-)
I will be closing this...