Proof-of-concept Genetic Algorithm Optimization Build

Check for duplicates

• I've checked for duplicate issues by using the search function of the issue tracker

Is your feature request related to a problem?

Yes.

Describe the solution you'd like

I did two separate projects. In the first Lua implementation of the UI and the implementation of the genetic algorithm itself in Rust.

https://github.com/Tiikara/path_of_building_genetic_genetic_lualib - genetic library on rust for lua
https://github.com/Tiikara/pathofbuildinggenetic - PoB with genetic

Describe alternatives you've considered

No response

Additional context

Currently implemented:

• Genetic algorithm
• Multithreaded calculation
• UI interface for setting the values the genetic algorithm will aim for
• Saving the genetic algorithm settings in Build xml
• Displaying the best results in the UI during the optimization process
• Use of conventional nodes
• Using ascendence nodes
• Using mastery nodes

Current issues:

• Slow performance due to long shortest path search (the algorithm takes 20 minutes on my Intel Core i9 12900k and uses 8 gigabytes of memory)
• Memory leak on optimization restart

What needs to be done before merge:

• My fork uses the 64bit architecture and luajit2.1 from https://github.com/openresty/luajit2. Can we already recompile the project to 64bit? :)

The 64bit architecture is necessary because of threads' memory usage. The luajit2.0 used in the project now cannot allocate more than 2 Gbytes of memory, that's why I chose https://github.com/openresty/luajit2 without this limitation.

Project dependencies (runtime_x64) https://github.com/Tiikara/pathofbuildinggenetic compiled with Visual Studio 2019

Start with the linked pathofbuildinggenetic fork and its runtime_x64 dependencies, then compare the Lua UI and Rust genetic library with the main project's LuaJIT 2.0 setup. Reproduce the optimization workflow and investigate the reported 20-minute runtime, high memory use, and restart leak. Done requires a defined merge plan for 64-bit support and agreement on the remaining performance and memory issues.