Copying Lyra's animation system - Getting numerous linked layer calls per frame??

Development Character & Animation
,
1

This question is aimed at anyone who has attempted to copy Lyra’s animation system into a standard UE project. And when I say “copy”, I mean actually recreating the system… NOT migrating files from Lyra. (So with that said, it’s going to be very hard for me to share code for troubleshooting (other than pieces of it.) If you’ve attempted this yourself, you know how involved it is.)

Here’s my rundown: Everything “works” in my game, to the point where stuff animates and appears to visually be correct. It looks visually fine.

However, I’m fighting against an annoying problem where the profiler is telling me that the Linked layer (i.e. ABP_PistolAnimLayers) is running on 1 character 17 times per frame. I started profiling, because even 20 npcs has extreme slowdown, when it clearly should not be (Lyra itself doesn’t suffer this much, and my old ABP was WAY faster.)

Yet, if I profile the real Lyra itself, it seems to show about 1 ABP_PistolAnimLayers count per frame per NPC active.

My first thought is, I have 2 child meshes that follow the leader mesh. Followers AREN’T supposed to run animation code tho, they’re supposed to just inherit the final results, right? So I don’t think that’s it. Also its only 2 skeletal children.

I was even thinking maybe theres something special in Lyra’s AnimInstance class. What is the difference between AnimInstance and LyraAnimInstance?

I’m also certain I’m not outputting anything weird in the ABP_Base itself. I even went as far as disabling the AnimGraph entirely on the base, and it still runs PistolLayer 17 times.

Does anyone with experience in this system, have any ideas of what I might have missed?
Any help would be appreciated!

2

Well, I think I cut the calls down to what looks somewhat normal.

Apparently you need to have all your ALI layers in 1 “group.” I noticed it after recreating a test ABP with a simply ALI, and my new test was doing basically 1:1.

I wouldn’t have thought that even mattered, especially when I saw people on the internet doing ungrouped. I guess as always, follow Epic’s demos first, instead of some rando’s video on the internet.

The performance still isn’t great, but that may be different than the topic I posted. Hopefully this reply helps some person googling this in the future.