Jan Wassenberg

> small array sizes are the base cases for large array sizes, and that vqsort may not perform as well as vxsort or djbsort on small arrays (and that this...

Thanks, that is summary is helpful for discussion. One thing we should clarify is that this plot is generated from a microbenchmark which is repeatedly called for the same length,...

Interesting. Can you share the numbers as a table or plot?

Thanks for sharing. FYI 94a07e8 is broken, we should have the fixed version up by tomorrow. Looks like your benchmark has a much lower peak after 2^7.

:) FYI the padding optimization has just landed.

I've added vxsort to our bench_sort. Results for AVX-512 and AVX2 with 100M keys are: ``` [ RUN ] BenchSortGroup/BenchSort.BenchAllSort/AVX3 AVX3: vxsort: u64: uniform32: 1.00E+08 426 MB/s ( 1 threads)...

Hi @sharpobject, FYI we have a [follow-up with new measurements](https://github.com/google/highway/blob/master/hwy/contrib/sort/README.md#additional-results) after fixing a performance bug, which was especially affecting short arrays. Happy to discuss if you're interested.

Closing, feel free to reopen if anyone wants to discuss.

Hi, thanks for reporting this. Interestingly, in Godbolt GCC 13.1 seems to work fine: https://gcc.godbolt.org/z/5jfdfn7ab Does anyone know if Graviton3 ships with a GCC tailored to what their core can...

Nice. It's probably helpful to use a newer GCC than 7.3, so perhaps one good way forward is to build GCC from source, or is there some other package repo...