How hard is it to have blender render tiles and continue to further frames without having to wait for the whole frame to finish to save render time?
A list of tiles rendered would be a great thing to have… It could also be used for network rendering if the list updates fast enough.
I always wonder why people keep asking “how hard” things are. What difference does that make? Are you going to feel better about not having this feature if I told you it’s incredibly hard? Are you gonna vehemently insist on having this feature ASAP if I told you it was dead simple? Or maybe it’s just a way to say “I would like to have this feature” without actually saying it.
For what it’s worth, if one was to implement this feature in a simple manner, the scene would have to be loaded into memory twice, there would have to be two destination buffers… roughly doubled memory usage then. The whole thing is rather pointless anyway, because the only way it would save render time is if a frame was stuck on less than N tiles in the previous frame (where N is the number of cores/GPUs in your system). If you believe there is something to be gained here, just render two instances of Blender on the commandline (which is a bit faster anyway).
How hard it is to answer “Even if it wasn’t hard, that doesn’t mean someone with sufficient knowledge would code it”?.
This way, the OP would know how blender works a little better, and you would have been a positive member of the community by teaching him in a positive manner.
How is that in any way positive or educational? I actually took the time to explain that the feature would bring no benefits that can’t already be taken advantage of by existing features. That’s educational!
This is a veiled feature request and I reserve the right to be negative, snarky and dismissive towards them! As for being a “positive member”: I obviously don’t care about that, but thanks for the free lecture.
@lordodin, funny i’ve been talking with someone who want to create that… so perhaps within a few weeks thats possible.
Its actually not a hard problem (pretty easy). split image, you only need a simple network master slave feeder service
So a slave can ask what tiles are left.
Deviding render into render borders isnt hard, there are a few scripts who allready do that.
I could code suc service in c# but … i’m not into python, so my solution would only work for windows (far from ideal).
But if you get the idea and are into python coding… it’s less then 1 hour coding.