(Question) Improve performance using parallelism

[jrodrigv]

Hi, this is just a question/brainstorm.

Is there room for performance improvement by means of creating a pipeline where modules/patches can be somehow processed in parallel?

Kind Regards

[blowfishpro]

Processing patches in parallel could be messy since certain patches may rely on other patches having been run before them. Maybe you could work something out here but it would be complicated.

What I am working on right now is for ModuleManager to run on a separate thread while KSP is still loading models and textures. I think this is somewhat safer since patches are still applied sequentially and they only need to be done before KSP starts compiling parts.

[jrodrigv]

Thanks for the answer @blowfishpro.
I think that should improve the performance.

I was thinking that it might be possible to perform a pre-processing of the patches in order to define a tree of dependencies and then process them in different threads but limiting the order between two specific patches if they depend on each other.

[blowfishpro]

One of the biggest complications is that there's actually a syntax for modifying things outside of the current node you're patching. Might be doable with some locking, but it would be pretty complicated to implement. There's still a lot of lower hanging fruit as far as optimization is concerned.