As you may know, I’m using Julia for performance/energy evaluation here GitHub - cecimerelo/BraveNewAlgorithm.jl I routinely test new versions for improvements…
After initially measuring a downgraded performance, I took other tests and there seems to be a slight, but significant improvement.
I have checked the diff and, although I’m no expert, there are some changes in memory usage, marking it as volatile, as well as some change in types in strings (from Integer to Int), changes in upstream versions of tools… I can’t pinpoint any specific thing that would have made it…
I think at a minimum it would be useful to understand what you are actually measuring here when you say “time and energy consumption”. If there are actual runtime regressions the next step would then be to try and isolate where in your benchmarking suite they occur.
After measuring again, there seems to be a slight, but significant improvement in, at least, energy consumption. Measuring performance is a complicated affair always… I’m editing the title to reflec that.