Shefki (co-author of PitBull4) writes:
State of PB4:
As of the push tonight I believe the major performance issues are taken care of and generally serious bugs are fixed. There are of course a few here and there and some feature enhancements I'd really like to do. But the point here is that I consider us pretty much to be pretty much to a beta state.
While I consider us to be at that point I won't be immediately tagging the project as a beta. I'm gonna let PB4 simmer for a little bit to see if some things come up. I figure it'll get tagged probably sometime this weekend.
All that being said I wanted to talk about performance. Particularly as to how it is going to impact this timeline. I'm very interested in performance problems. PitBull in the past has had adequate performance, but it still has been somewhat lacking. I think we can provide the configuration options that our picky users want while still maintaining top notch performance.
As of the changes tonight during a typical AV run PB3+DogTags will use about 3.6 seconds of CPU time per minute of battleground time. PB4 using LuaTexts is using about 2.5 seconds of CPU time per minute of battleground time. PB4+DogTags is using about 6.4 seconds of CPU time per minute of battleground time.
The above is with roughly identical configurations between the 3 different setups. The DogTag performance numbers is including the time the LibDogTags-3.0 and LibDogTags-Unit-3.0 libraries took (PitBull was the only addon using them).
I strongly recommend that people still using DogTags and having performance problems migrate to LuaTexts. When reporting performance problems please do say if you're using DogTagTexts or LuaTexts, as you can see there is a drastic performance difference.
I have a pretty good idea as to what is causing the performance problems with DogTagTexts. Unfortunately, I don't have a very good solution for the problem. DogTags was largely written with PitBull3 and CowTip in mind. PitBull3 puts a lot of work on the individual modules to know when to update frames, particularly frames that we do not receive events for. PB4 includes this work in the core. Unfortunately, DogTags takes on this work and it gets doubled up. This explains the performance numbers seen above as the PB4 number is slightly less than double the PB3 DogTag number.
As a result of this it's almost guaranteed that I'll be switching the default text provider to LuaTexts.
I want to be clear that I'm not giving up on fixing DogTagsTexts. However, it was always the intention of making a lighter weight text provider be the default. I'm going to encourage people to shift to LuaTexts pretty strongly. I hope this post helps people make the decision to switch.
I also hope it demonstrates that PB4 is at this point outperforming PB3 with the exception of DogTags. I look forward to your feedback.
This thread is intended to work as http://elitistjerks....ags_3_0_thread/, but obviously for LuaTexts instead. Unsure how to get started? Check this out.