This might have been suggested before, but I'm not sure.
Currently when doing Distributed Rendering using LightCache, each machine involved does it's own LightCache calculations. On my network at least, the workstation I'm submitting the DR from is much faster than my rendernodes, and so completes the LightCache pass well before the rendernodes.
A way to speed things up would be if the first machine to finish the LightCache calcs could Flag that it was done, and send its LC solution to all the other machines (of course this probably means sending it back to the workstation first).
This thought occured to me before, but really emphasized it as I'm sitting here watching this rendering. It's a high-resolution still, and my workstation has gotten through about half of the IRRMap solution, and my 10 render slaves haven't even finished their LC passes yet.
While I know people have asked for distributed LC calcs in the past, and that is a very difficult thing to do, I think this would be a great intermediate solution to help speed things up.
Currently when doing Distributed Rendering using LightCache, each machine involved does it's own LightCache calculations. On my network at least, the workstation I'm submitting the DR from is much faster than my rendernodes, and so completes the LightCache pass well before the rendernodes.
A way to speed things up would be if the first machine to finish the LightCache calcs could Flag that it was done, and send its LC solution to all the other machines (of course this probably means sending it back to the workstation first).
This thought occured to me before, but really emphasized it as I'm sitting here watching this rendering. It's a high-resolution still, and my workstation has gotten through about half of the IRRMap solution, and my 10 render slaves haven't even finished their LC passes yet.
While I know people have asked for distributed LC calcs in the past, and that is a very difficult thing to do, I think this would be a great intermediate solution to help speed things up.
Comment