Hacker News Reader

I can imagine the overhead for ints is large. But how often do you sort pure ints? A more realistic test would be some objects sorted on a field value.

View on HN

I've updated the repo and README.md to reflect this case. For sorting a struct with (int,string) on it's random integer value, I got the following:

[10K elements] BenchmarkLibPotato-4 10000 2166569 ns/op

BenchmarkSpecPotato-4 10000 1020842 ns/op

[10M elements] BenchmarkLibPotato-4 3 3829722562 ns/op

BenchmarkSpecPotato-4 10 1802874882 ns/op

That's great feedback. I'll try out this specialization on a simple struct (int32/string) combo later today. We can still get rid of the indirections around slice manipulation (Swap and Len) but obviously stay stuck with the Less.

Or try sorting points (struct of two float64s) by the x-coordinate, and compare the timing with similar JavaScript code running in V8. You might be surprised...

Reply on news.ycombinator.com