cbrt
cbrt
This is something I desperately want to see in a musicbrainz tagger and since picard doesn't have this option either I would be more than happy to have this option
The benchmarks of `Sorted` are sketchy at times with there occasionally being outliers, but I think overall there isn't a noticeable difference - when you consider the code that should...
mentioning @Bodigrim as they originally told me to come here from the [original discussion](https://gitlab.haskell.org/ghc/ghc/-/issues/24280)
I do not have the technical / low level knowledge to write such a benchmark but I would welcome someone contributing it if it's a concern
@Bodigrim I think it would make more sense to add some specific cases to test correctness and stability (I'm thinking [], [x] and one with a few more) As far...
correction: there are tests under `libraries/base/tests` but not under `testsuite/tests/lib/base`
I have added a benchmark for number of comparisons to my repo. @int-e's 4-way merge consistently gets less allocations and comparisons and also performs best when considering CPU time. (Allocations...
I think the version from @int-e is the best way to go forwards now. It's better across the entire board and there are no tradeoffs. If anyone wants to recreate...
> I remain concerned about the performance risks of such a heavy `sort` in complex contexts. The code size of a four-way sort is nontrivial; is it enough to cause...
> One possibility is to add this to `Data.List` as a _new_ function, rather than replacing the existing `sort`. Then users can choose to adopt it if it meets their...