News for hackers. Stuff that matters.
Re:l: A new runtime for k and q by jloveless on Tuesday July 07, 2026 @04:24PM (#48823205)
Re:l: A new runtime for k and q by jazza68 on Tuesday July 07, 2026 @05:14PM (#48823887)
that is fast! Are you going to do the STAC-M3 benchmarks? Kx use them as evidence that they are the fastest, which is a fair claim as nobody else participates in that particular test :-)
Re:l: A new runtime for k and q by jloveless on Tuesday July 07, 2026 @05:16PM (#48823920)
I think one day we'll do that on https://oxide.computer/ 1: because we love our friends 2: the gear is great (and ROCEv2 supported which L uses)
Re:l: A new runtime for k and q by kristjansson on Tuesday July 07, 2026 @05:02PM (#48823702)
ah. cheeky.
Re:l: A new runtime for k and q by kristjansson on Wednesday July 08, 2026 @10:49AM (#48832730)
Seems like the only point of comparison against a competitive k/q runtime is an MBP run? suggestive to be sure, but would be good to see some other machines, esp. an x86.

Seems wild that the much-lauded "reference engine" leaves ~2x perf improvements on the table.

Re:l: A new runtime for k and q by jloveless on Wednesday July 08, 2026 @11:23AM (#48833173)
I’d encourage running the master bench yourself! Super easy

https://github.com/l-labs/master-benchmark

Or on the database side of the house try the h2o.ai base that duckdb excels at

https://github.com/l-labs/db-benchmark/tree/main/l