This makes the node system somewhat more high-level with the intent of making
working with them far more flexible and stable. By making the architecture more
abstracted, it becomes far less rigid which should allow us to do even more
with it and make it much less crash prone.
Yep, this is another one of my "famous" sweeping rewrites. Expect things to break.
Goals for this are:
- Greatly simplify node connections (particularly with arrays) so the code requires less maintenance/is more stable
- Redesign node structure to address issues where UI would stall for lengthy periods of time
- Less reliance on shared ptrs/greater reliance on QObject system for inheritance/memory management
- General code cleanup and improvements
The macro defined namespaces confused the hell out of lupdate and more or less broke translations permanently. Looks like the only way we can do it is to have a hardcoded namespace, which goes against my instinct, but honestly how likely is it that we'll change the namespace anyway (I guess forks might want to do it, but that's their problem ;) )
The nodes now have more control over how their accelerated shaders/sample
functions are run, as well as how items are popped off the value tables.
This allows for various optimizations that we didn't have access to before.
Since we load in a separate thread, when the QObject based objects are
instantiated, they're created with affinity to that separate thread. Now we
specifically ensure they are moved to the main thread after their creation.
Previous iteration used some "magic code" that added clips automatically to the
timeline. This was functional but ultimately outside of the undo commands'
control meaning nodes could be infinitely added and abandoned. This makes the
add process part of the undo command which means it's all undoable as the user
would expect.
If the nodes are now stateless, there's nothing stopping the renderer from
rendering multiple frames at once. Earlier since the nodes held some of their
input/output data (and that data could change per frame), it was not possible
to render multiple frames at once without conflicts. Now that the node state is
held in render threads, they can do whatever they want at any time.
If a NodeParam is destroyed (presumably as the result of its parent Node being
destroyed), any edge objects connecting to another param should be destroyed
or else the edge will be floating referencing a now destroyed object.
Old code assumes that a NodeParam's parent will always be a Node. The function
has been separated off and tweaked in the event that this is not the case.
Previous iterations would use mutexes to prevent changing of the graph
mid-render, however several user actions would need to capture these mutexes
causing the main thread to hang until the current render job (frame/range of
samples) was complete. We now copy the nodes necessary as part of the "compile"
process so that the main thread shouldn't need nearly as much blocking while
caching occurs.
commit
For testing the new iteration, the texture cache disk download was written
into the main thread instead of into the separate threads. Now they're back
in separate threads again.
Also I think some of these files probably should have been in the previous
commit.
The new rendering pipeline strives to simplify the nodes themselves as much
as possible and move much of the logic to an external rendering engine. This
change removes all of the responsibilities that no longer belong to the
nodes themselves and will soon be folded into the renderer.
Nodes were previously written to be "strongly typed" in that a parameter's
"type" enforced whether it could be connected to another. All code related
to that has now been removed since not only is it hard to maintain and
likely unnecessary, it's possible the nodes will work differently later on
anyway.
Mike Solar