The devs. ultimately want the modifier system to be node based, so not having too many modifiers added at the moment will decrease the amount of work needed. An example of this is the planned refactor of the whole system for 2.8, it will still be stack-based for now, but the underlying code will be changed to make it ready for its conversion to nodes when the time comes.
I don’t think that somebody that worked in the blender foundation offices is not a main developer…
The on-hold: “Everything-Nodes” project.
Every time I read that, I see it as meaning “me specifically talking about animation with armatures” but I’m just now starting to think you were referring to the numbers I gave. 750k+ faces is animating an “object” in “object mode”. “far less faces” meant something more like 50k-200k faces, not something like 500k+ faces.
Speaking of which, I found that Blender’s peak spot for peak mesh animation performance (just rig, no textures) is around 50k faces (total in view), and Quads play around 25% faster than Triangles (because it sees 2 triangles as 2 faces, and 1 quad as 1 face, or because it’s counting less edges).
I remember reading (somewhere) that the new dependency graph was mostly finished, and the thing users are waiting for is copy-on-write, and that’s already available but “unstable” (as a startup command).
Would --enable-copy-on-write give an idea of what’s expected for 2.8, or should the results be seen more of as a quick-plug-backtrack to just “see” things?
Like should I give up hope (excluding any future updates) that the performance will be any different, or should I view it as a “legacy mode” and expect the “new/standard mode” to be something else that’s not available?
I ask because I download 2.8 weekly, hoping the performance would increase but always get practically the same fps as 2.7 (Clay render because Eevee mostly lags like Blender Render GLSL).
It is impossible to predict how the dependency graph is going to impact every case. What can be said is that it should in general be faster. But you may also encounter cases where this is not the case.
I don’t know whether the dependency graph is finished for 2.8 or if it is going to get another iteration of optimizations. You would need to ask the developers and Sergey in particular to get that information.
Unless they probably did not advance as fast as they thought and 2.79a was not released when they wanted, there is hope.
If you take a look to last commits relative to depsgraph that happened last week ; you can see things relative to camera animation, bones selection, shapekeys …
And at the same time, you can see Campbell trying to deal with the idea of a mode per workspace and tweaking object mode.
If hard core of depsgraph is almost finished ; obviously, connections to it is taking time to be made correctly for each thing.
And it is not the best moment to evaluate object mode.
If bugtracker is not yet open to users ; in a sense, it means that they are still trying to validate concepts, note what problematic is solved and which one is still unsolved and continue to search solutions by themselves.
They will continue to do that until beta release when user feedback would be legitimate.
Until that moment, their priorities are not necessary matching user ones and may seem obscure to users.
It may be frustrating to wait second quarter of 2018 to address a specific issue.
But we can not deny that many things are progressing. And that will continue.