- It uses same blend set up as blend list
- Because of this, now we're not testing weight <ZERO_ANIMWEIGHT, but we test Blend.IsComplete
- Now we still advance montage since if the custom curve contains longer 0, anim is likely to pause if he comes back from 0. (I had wacky test case)
#ANIM: Single node instance's inconsistency of montage weight has been resolved
- This also contains montage update change where now montage advance prior to node update
- Also made new function for node update, so that single node instance can use that function instead of nativeupdateanimation, which happens prior to weight update causing all inconsistency
- Enabled node weight debuging define for montage, so that we can see if any other case is causing it
- This also fixes crash when montage gets terminated after weight has been updated
Merging
//depot/UE4-Orion/...
to //depot/UE4-Orion/...
[CL 2687868 by Lina Halper in Main branch]
- This checkin eliminates use of all <Module>Classes.h files in Engine, with the exception of UnrealEdClasses.h
- Compilation speed-ups for all of the modules affected, now that they are including what they use
- No effective changes other than C++ includes
[CL 2686054 by Mike Fricker in Main branch]
- at the end, no change on the data because this guarantees frame independent.
- Added help and renamed variable to speed, so that it knows it is speed.
Merging
//depot/UE4-Orion/Engine/...
to //depot/UE4-Orion/Engine/...
[CL 2683714 by Lina Halper in Main branch]
- Now it gets weight data and use that weight data without recalc in eval
- Also made montage instance's members to be private
- changed single node to get weight and use that weight for eval - same as normal anim BP
- changed update weight of montage to happen before native update animation
Merging
//depot/UE4-Orion/Engine/Source/...
to //depot/UE4/Engine/Source/...
[CL 2682922 by Lina Halper in Main branch]
Merging
//depot/UE4-Orion/Engine/Source/Runtime/Engine/Private/Animation/AnimNode_Slot.cpp
to //depot/UE4/Engine/Source/Runtime/AnimGraphRuntime/Private/AnimNodes/AnimNode_Slot.cpp
[CL 2682911 by Lina Halper in Main branch]
Optimization is enabled by default but is an engine setting in case it needs to be disabled (bOptimizeAnimBlueprintMemberVariableAccess).
Expanded FExposedValueHandler to also contain an array of source & dest properties (copy records) to copy between. These are resolved to read/write addresses on init & a simple memcpy at runtime instead of calling the usual generated custom event.
Custom events are not added at all if all properties use copy records. The event may still be added & called however as mixed-mode access is still supported (i.e. a thunk and a memcpy to different pin properties).
Unfortunately this required initialization of all ExposedValueHandlers (as I didnt want to add the branch overhead of lazy init) so all the anim nodes have needed a small change to call their base class Initialize().
[CL 2678504 by Thomas Sarkanen in Main branch]
//depot/UE4-Orion/Engine/Source/Runtime/AnimGraphRuntime/Private/BoneControllers/AnimNode_Trail.cpp
to //depot/UE4/Engine/Source/Runtime/AnimGraphRuntime/Private/BoneControllers/AnimNode_Trail.cpp
[CL 2671622 by Lina Halper in Main branch]
//depot/UE4-Orion/...
to //depot/UE4/...
UE4 - rollup of many performance changes
* increased max threads for task graph and added a way to reduce task threads at runtime for testing the optimal number
* added a way to track excessive allocation at runtime without using the memory profiler: LogGameThreadMallocChurn.Enable
* ps4: fixed a few problems with the platform properties, IsRunningGame was returning false
* added inline storage for one delegate in multicast delegates
* fixed excessive allocation on the script stack traceback
* fixed excessive malloc calls throughout the animation system
* added stats throughout the animation system
* reworked parallel queue ticks
* fixed excessive malloc calls in cloth and physics animation systems
* fixed GHitchThresholdCVar
* minor improvements to abtest
* added ability to change thread affinities on the fly so they can be tuned
* cvars to control animrate optimization and spew the results: EnableAnimRateOptimization, SpewAnimRateOptimization, DrawAnimRateOptimization, ForceAnimRate
* optimized malloc calls in cascade and gpu particles
* optimized malloc calls in input system
* removed visual logging and ability log from consoles and fixed a few compile errors related to doing that
* optimized gameplaytags
* added a few cvars to cause stalls in various parts of the pipeline to identify critical paths and bottlenecks: CriticalPathStall.*
* added console command to toggle all known outstanding perf tweaks: Orion.SetupPerfCVars
[CL 2642061 by Gil Gribb in Main branch]
- One driver value can now drive multiple components
- How a component is driven is now configurable (add to input, replace, add to ref pose and replace)
- The range limit is now defined on the input (before the multiplier is applied) rather than the output (existing ranges are fixed up on load)
- New details customization to make working with the node easier
- Individual scale components can now be a driver or one of the driven components, rather than just the max component of scale
- The node title now displays the driving equation (e.g., driven.rotation.z = -0.5 * driver.rotation.x)
Animation: Added a custom version for AnimGraph/AnimGraphRuntime usage
Upgrade Note: The exact behavior of the original node was not preserved, but there are now more configuration options, so please revisit any uses of bone driven controllers in your animation blueprints
[CL 2606968 by Michael Noland in Main branch]
