Why is it that the animation complexity is bounded by the memory on each machine? Are the communication costs too large to distribute it across multiple machines?
This comment was marked helpful 0 times.
rbcarlso
If an animator wants to see motion in real time, yes. It depends on how long a delay you want between specifying the animation, rendering, and output.
This comment was marked helpful 0 times.
spilledmilk
I believe the animation complexity is bounded by the memory on all the machines. The processing is distributed across all the machines, but the maximum number of polygons that can be rendered is still bounded by the total memory available on all the farm machines.
This comment was marked helpful 0 times.
vrkrishn
Im confused about what the last bullet point is saying, specifically what the footprint refers to. Does this mean that the total memory of the system is constrained but if you add more processors, there has to be more memory allocated to managing processor interactions?
This comment was marked helpful 0 times.
pwei
@vrkrishn I'm thinking the same thing as you are. I believe that in general, the more processors you are using, the more overhead there is to manage them. One naive example that I can imagine (which I might be completely wrong about) is that if we're rendering a scene, and we want it raytraced (a common option in final renders), then splitting the scene into layers will require that each processor have information about the above and below layers in order to carry out its own computation. In that case, keeping that information handy can create this memory overhead, as rendering sequentially can just keep all the memory in one place.
Why is it that the animation complexity is bounded by the memory on each machine? Are the communication costs too large to distribute it across multiple machines?
This comment was marked helpful 0 times.
If an animator wants to see motion in real time, yes. It depends on how long a delay you want between specifying the animation, rendering, and output.
This comment was marked helpful 0 times.
I believe the animation complexity is bounded by the memory on all the machines. The processing is distributed across all the machines, but the maximum number of polygons that can be rendered is still bounded by the total memory available on all the farm machines.
This comment was marked helpful 0 times.
Im confused about what the last bullet point is saying, specifically what the footprint refers to. Does this mean that the total memory of the system is constrained but if you add more processors, there has to be more memory allocated to managing processor interactions?
This comment was marked helpful 0 times.
@vrkrishn I'm thinking the same thing as you are. I believe that in general, the more processors you are using, the more overhead there is to manage them. One naive example that I can imagine (which I might be completely wrong about) is that if we're rendering a scene, and we want it raytraced (a common option in final renders), then splitting the scene into layers will require that each processor have information about the above and below layers in order to carry out its own computation. In that case, keeping that information handy can create this memory overhead, as rendering sequentially can just keep all the memory in one place.
This comment was marked helpful 0 times.