Profile and optimize implementation of GEKPLS

[vikram-s-narayan]

Profile implementation of GEKPLS and analyze and fix bottlenecks.

[Spinachboul]

Tasks done

I profiled the function GEKPLS and identified the potential bottlenecks, which might be causing heavy memory usage:

• abstractinterpretation.jl
• typeinfer.jl
• boot.jl

And upon individual inspection, I discovered that the function _ge_compute_pls might be the most heavily resourceful function for which I will provide the profiling information soon.

@ChrisRackauckas & @vikram-s-narayan
Please look into the flame graph and suggest any possible changes!!

Here is the link to the notebook: https://julianotebook.w3spaces-preview.com/index.html

[ChrisRackauckas]

It looks like you included compilation in the profiling.

[Spinachboul]

Yeah! I just tried to profile the whole function to get a high-level view of memory storage and time consumption. I am working on gekpls.jl where I would suggest a suitable sampling technique for based on rmse scores for different sampling methods.

It looks like you included compilation in the profiling.