https://benknoble.github.io/blog/2024/02/15/benchmarking-pict-equality/
I'm sharing this here partly for general discussion and partly because folks might be interested. I don't entirely understand the results, so I'm open to speculation or analysis on that front.
P.S. Getting the code right for the GC in memory benchmarks made a big difference! (See the history of the memory.svg plots in the repo.)