Marshall Lochbaum

Nice! I [computed tables](https://mlochbaum.github.io/BQN/try.html#code=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) of failure probability for median combinations. They're shown as negative log base two of the probability, so 20 is roughly one in a million and so...

I don't get the concern about randomization breaking detection of bad patterns. The only thing that can slow down a branchless quicksort is bad pivot selection, right? But what qualifies...

I rounded up the number of new candidates to a power of two if there's not enough swap space for a normal merge, and everything seems to work now. There's...

Comparing to older versions, I find that after some tuning it's about the same as the version from before I started saving candidates. Both definitely better than the version I...

In a crazy coincidence it seems like √n is pretty much exactly the number of candidates needed to distinguish a bad input from a uniform random one in Robin Hood...

Did more work with the RH criterion. The candidate scanning does seem to work well, but it obviously has no power when there's only one candidate selected by pseudomedian of...

Wrote [this section](https://mlochbaum.github.io/BQN/implementation/primitive/sort.html#candidate-selection) on techniques with candidate selection. The last paragraph discusses choosing merging rather than partitioning for one step, which is a new idea (to me at least). It's...

The proposal part is very similar to what I tried to do with [this post](https://topanswers.xyz/apl?q=1908). Although some design discussions on TopAnswers have been successful, I haven't been able to get...

My comments after looking closer at SRFI. It's an interesting system, thanks for pointing it out! The process seems to be heavier than we'd want at least initially, with the...

Oh, I can barely stand Rust. At least they have a benchmark I can run, but I'm getting wildly different numbers from theirs. ~80% CPU usage on 2 cores /...